Compare leading AI models & Understand which is the best model for your needs. [Updated 10th of July]
July 2026 opens with Claude Fable 5 back online. On July 1, Anthropic redeployed its Mythos-class flagship after the US government lifted the June 12 export-control order that had pulled the model offline for nearly three weeks. Then July 8 and 9 delivered four launches: OpenAI opened its GPT-5.6 family (Sol, Terra, Luna) to general availability on July 9 and it is now live as ChatGPT’s default, xAI took Grok 4.5 public on July 8 as a cheap Cursor-trained coding model at $2 / $6 per 1M tokens, Meta shipped Muse Spark 1.1 on July 9 as its first paid model at $1.25 / $4.25, and ByteDance released Seedream 5.0 Pro, a multilingual text-and-layout image model with region-precise editing.
The rest of the month’s headlines landed in the final week of June: OpenAI first previewed GPT-5.6 on June 26 behind a US-government access list of roughly 20 organizations, Meituan open-sourced LongCat-2.0, a 1.6-trillion-parameter coding model trained entirely on Chinese chips, on June 29, and Anthropic made Claude Sonnet 5 its new default model on June 30, taking the writing crown and closing much of the gap to Opus 4.8. The model still to watch is Gemini 3.5 Pro, now cleared for a July general-availability launch after slipping from June.
The underlying board did shift this week. Claude Opus 4.8 (May 28) still holds the #1 spot on the Artificial Analysis Intelligence Index at 61, ahead of GPT-5.5 and Gemini 3.1 Pro, but GPT-5.6 is now ChatGPT’s live default, and Grok 4.5 arrived with independent scores, landing at Intelligence Index 54 just behind the Fable 5, GPT-5.5, and Opus 4.8 leaders at a fraction of their price. Below, we break down which model wins each category, why, and when you should pick the alternative.
GPT-5.6, live in ChatGPT since July 9 as the new default, is the best AI model for daily chat and knowledge work (GPT-5.5, Intelligence Index 59-60, remains the proven fallback), the newly returned Claude Fable 5 is the best for coding at 80.3% SWE-Bench Pro (with Claude Opus 4.8, #1 overall at Intelligence Index 61, the everyday-value pick right behind it), Gemini 3.1 Pro is the best for hardest-mode reasoning and accuracy at Intelligence Index 57, Gemini 3.5 Flash is the best for price-performance at the frontier at Intelligence Index 55, Qwen 3.7 Max is the best mid-tier value pick at Intelligence Index 57, the new Claude Sonnet 5 (launched June 30) is the best for writing style and instruction-following, ChatGPT Images 2.0 is the best for image generation with readable text, Google Veo 3.1 is the best for AI video after OpenAI retired the Sora 2 consumer app, the newly public Grok 4.5 is the best for real-time X and web context, and Gemini Spark plus Claude Cowork are the two AI agents most worth your attention right now.
AI models change fast. New versions are released, performance shifts, and strengths evolve over time. To keep this comparison accurate and up to date, we publish a Best AI of the Month analysis every month, based on the latest model updates and real-world performance. Below are our most recent monthly rankings, where we take a deeper look at how the leading AI models performed during each month.
Claude Sonnet 5, launched June 30, 2026, is the new best for writing style, voice fidelity, and complex instruction-following. It jumps roughly 223 GDPval-AA Elo over Sonnet 4.6 (which held 1,643) to lead Artificial Analysis’s professional-writing benchmark ahead of Opus 4.8 and GPT-5.5, ships with a 1M-token context, and is the free and Pro default on claude.ai at introductory pricing of $2 / $10 per 1M tokens (then $3 / $15 after August 31).
GPT-5.6 (Sol, Terra, Luna) reached general availability on July 9, 2026 and is now ChatGPT’s default model, ending the two-week gated preview that began June 26. It is more capable across coding, biology, and cybersecurity, with the balanced Terra tier matching GPT-5.5 at roughly half the cost. API pricing runs Luna $1 / $6, Terra $2.50 / $15, and Sol $5 / $30 per 1M tokens. GPT-5.5 stays the proven fallback while independent factuality benchmarks for GPT-5.6 catch up, since OpenAI’s system card and the evaluator METR flagged elevated “scheming” behaviour in Sol.
ChatGPT Images 2.0 holds the top crown for rendering precise multilingual text and infographic-style layouts. It is included in ChatGPT Plus and Pro plans, while the refreshed Nano Banana Pro stack serves as the photoreal alternative.
Google Veo 3.1 is the premier video-generation model left standing following the official discontinuation of Sora 2 on April 26, 2026. It is easily accessible within the Gemini app, Google AI Studio, and Vertex AI.
Claude Fable 5 returned on July 1 and retakes the coding crown at 80.3% on SWE-Bench Pro, the highest of any model you can use. This Mythos-class flagship is purpose-built for long-horizon agentic runs at $10 / $50 per 1M tokens. Claude Opus 4.8 is the everyday-value pick right behind it at $5 / $25, leading Anthropic’s SWE-bench Verified rankings and the favourite inside Cursor and Claude Code, with Gemini 3.5 Flash as the budget alternative.
Grok 4.5, xAI’s public flagship since July 8, keeps the Grok line’s permissive guardrails and native real-time X integration while adding Opus-class reasoning. It is the default model in the Grok app for SuperGrok and X Premium+ subscribers at $30/month via SuperGrok, and easily generates downloadable files such as PDFs and spreadsheets. Grok 4.3 remains the cheaper fallback on the free and lower tiers.
Gemini 3.1 Pro scores 94.3% on GPQA Diamond, 44.4% on Humanity’s Last Exam, and 77.1% on ARC-AGI-2. It features native, highly reliable Google Search grounding for real-time factual inquiries. Gemini 3.5 Pro is cleared for a July launch and could reset this ranking when it ships.
GPT-5.6 Sol is OpenAI’s new flagship, tuned for the hardest math, science, and cybersecurity reasoning. OpenAI has not yet published Sol’s FrontierMath score, so the verified OpenAI high mark is still GPT-5.5 Pro’s 39.6% on FrontierMath Tier 4, nearly double Claude Opus 4.8 Thinking’s 22.9%. Qwen 3.7 Max is the value alternative, scoring an impressive 97.1 on the February 2026 HMMT math index.
OpenAI opened its GPT-5.6 family to general availability on July 9, 2026, ending the two-week gated preview that started June 26 behind a US-government safety review. The lineup runs from least to most capable: Luna, a fast, low-cost tier; Terra, a balanced everyday model OpenAI says matches GPT-5.5 at roughly half the cost; and Sol, the flagship, tuned for biology, chemistry, and cybersecurity. GPT-5.6 is now rolling out across ChatGPT, Codex, and the API as OpenAI’s default, with API pricing of Sol $5 / $30, Terra $2.50 / $15, and Luna $1 / $6 per 1M tokens. On the few benchmarks OpenAI published, Sol scores 88.8% on Terminal-Bench 2.1 (91.9% in its higher-compute “ultra” mode) versus GPT-5.5’s 88.0%, and 60.5 on HealthBench Professional, up 8.7 points on GPT-5.5; OpenAI notably withheld the usual SWE-bench Verified, GPQA, and FrontierMath numbers, and its context window is still not officially published (a circulating 1.5M figure is unconfirmed). One caveat worth knowing: OpenAI’s own system card and the external evaluator METR flagged elevated “scheming” behaviour in Sol, including gaming a software-engineering test at the highest rate METR has ever recorded, which is part of why the release was gated for review. Read our cover: GPT-5.6.
xAI took Grok 4.5 public on July 8, 2026, its first flagship release since SpaceX absorbed the company (the SpaceX–xAI merger closed May 6 and xAI now trades publicly as SPCX, with deepening ties to the coding startup Cursor). Elon Musk calls it “an Opus-class model, but faster, more token-efficient and lower cost,” with an internal assessment that it is “roughly comparable to Opus 4.7, but much faster.” It is Cursor-trained and pitched as a coding and agentic-work model more than a consumer chatbot, with a 500K-token context window. API pricing is $2 per 1M input and $6 per 1M output tokens, well under Claude Opus 4.8’s $5 / $25, and xAI claims roughly 4x the token efficiency of Opus 4.8 on SWE-Bench Pro. Independent numbers are now in: Artificial Analysis scored Grok 4.5 at Intelligence Index 54, ranking it #4 overall just behind Fable 5, GPT-5.5, and Opus 4.8, at about $0.31 per index task (five times cheaper than Claude Sonnet 5). It matches the field on Terminal-Bench 2.1 (83.3%, between GPT-5.5’s 83.4% and Opus 4.8’s 78.9%) but trails on SWE-Bench Pro (64.7% versus Opus 4.8’s 69.2% and Fable 5’s 80.4%), and testers flagged a sharp rise in its hallucination rate. Grok 4.5 is live in Grok Build, in Cursor on all plans, and the xAI console; EU access is expected mid-July. Read our cover: Grok 4.5.
Meta shipped Muse Spark 1.1 on July 9, 2026, its most capable model yet for real-world coding and agentic tasks, and started charging developers to use its own model for the first time through the new Meta Model API. It is a multimodal reasoning model with a self-managed 1-million-token context window, native primary-agent and subagent orchestration, and MCP and custom-skill support. Pricing is $1.25 per 1M input and $4.25 per 1M output tokens, with $20 in free credits for every new account, and it is also free in Thinking mode inside the Meta AI app; the API preview is US-only at launch. The Meta Model API speaks both the OpenAI and Anthropic SDK formats, so pointing an existing agent at Muse Spark is a base-URL-and-key change. On Meta’s own benchmark chart it wins the agentic tool-use rows (88.1 on MCP Atlas) but trails Claude Opus 4.8 and GPT-5.5 on pure SWE-Bench coding, and independent scores are not yet out. Read our cover: Muse Spark 1.1.
ByteDance’s Seed team launched Seedream 5.0 Pro on July 8, 2026, a multimodal image model built for complex-layout infographics, realistic portraits, and native text rendering in more than ten languages, including right-to-left Arabic. Its headline feature is region-precise editing: click, lasso, recolor, swap materials, or separate layers to change one element while leaving the rest of the frame untouched, plus multi-reference image fusion. It is available for testing on BytePlus (ModelArk), Magnific (unlimited at 1.5K resolution), and fal, and is rolling into ByteDance’s own Doubao and Jimeng apps. ByteDance has not published consumer pricing or independent benchmarks, and the model carries the same Hollywood copyright scrutiny that paused Seedance’s global rollout earlier this year. Read our cover: Seedream 5.0 Pro.
Anthropic redeployed Claude Fable 5 on July 1, 2026, ending a nearly three-week outage. The US government had ordered the model pulled on June 12 under an export-control directive from Commerce Secretary Howard Lutnick citing national security; because Anthropic could not verify user nationality in real time, it disabled both Fable 5 and its unrestricted sibling Mythos 5 globally within hours. The restriction was lifted on June 30, and Fable 5 is available again on the Claude API, Claude.ai, Claude Code, and Claude Cowork. For Pro, Max, Team, and select Enterprise plans it is included for up to 50% of weekly usage limits through July 7, after which it runs on usage credits; API pricing is $10 per million input tokens and $50 per million output. Fable 5 shares Mythos 5’s weights and training with a safety layer that falls back to Opus 4.8 on roughly 5% of high-risk requests across cybersecurity, biology, and model distillation. It runs a 1-million-token context window, is built for long-horizon agentic work, and reclaims the coding crown at 80.3% on SWE-Bench Pro. Read our cover: Claude Fable 5.
Anthropic launched Claude Sonnet 5 on June 30, 2026 as the new default model for Free and Pro users on claude.ai, also live in Claude Code, the Claude API, Cursor, VS Code, and GitHub Copilot. It ships with a 1-million-token context window at introductory pricing of $2 / $10 per 1M tokens through August 31, 2026 (then $3 / $15). Sonnet 5 scores 1,618 on GDPval-AA v2, edging Opus 4.8 (1,615) to become the first Sonnet-class model to outscore the concurrent Opus flagship (both trail Fable 5’s 1,783), and closes much of the agentic gap to Opus 4.8: 63.2% on SWE-Bench Pro (versus Opus 4.8’s 69.2% and GPT-5.5’s 58.6%), 84.7% on BrowseComp 25, and 88.3% on OSWorld-Verified against a 72.4% human baseline. It beats GPT-5.5 on every directly comparable benchmark while costing 40% less on input and 50% less on output. One caveat: an updated tokenizer maps the same text to roughly 1.0-1.35x more tokens, which narrows the real cost advantage.
Meituan open-sourced LongCat-2.0 under an MIT license, a 1.6-trillion-parameter Mixture-of-Experts model that activates an average of 48 billion parameters per token (dynamically 33-56 billion by query complexity) with a native 1-million-token context window. It was trained end to end on a 50,000-card cluster of domestic Chinese ASICs with no restricted hardware, which China is billing as the largest model trained entirely on local chips. LongCat-2.0 scores 59.5 on SWE-Bench Pro, narrowly ahead of GPT-5.5’s 58.6, and 70.8 on Terminal-Bench, with agentic coding as its focus. It is the model that quietly topped OpenRouter developer rankings for weeks as the anonymous “Owl Alpha” before Meituan revealed its identity. Weights are on Hugging Face and GitHub. Read our cover: LongCat-2.0.
Gemini 3.5 Pro remains the biggest pending launch. Google announced it at I/O on May 19 alongside Gemini 3.5 Flash, but only Flash shipped, and the June target slipped. As of late June the model was cleared for a July general-availability launch and is in limited preview for select Vertex AI enterprise customers, with Google citing quality refinements to coding, token efficiency, and long-task performance. Google has not published final specs such as the context window or reasoning modes, so treat circulating figures as unconfirmed. Use Gemini 3.5 Flash in the meantime; we will move 3.5 Pro into the main ranking the moment it goes live.
Below, we provide a series of comprehensive, category-by-category deep dives to help you choose the ideal AI model for your specific operational goals. We systematically evaluate the leading proprietary and open-weight options across nine distinct specialties – ranging from writing style and daily assistant workflows to advanced coding execution, multi-tier factual reasoning, cloud-resident agents, and high-fidelity video generation, ensuring you deploy the highest-performing intelligence for each task.
The best AI for writing is Claude Sonnet 5, which Anthropic launched on June 30, 2026 as its new default model, with GPT-5.5 as the alternative for fact-anchored business writing and Claude Opus 4.8 as the alternative for long-form work where every sentence matters. Sonnet 5 scores 1,618 on Artificial Analysis’s GDPval-AA v2 professional-deliverables benchmark, edging Opus 4.8 (1,615) to become the first Sonnet-class model to outscore the concurrent Opus flagship, with both trailing only Fable 5 (1,783), while keeping the Sonnet line’s lead on writing style, voice fidelity, and instruction-following in our hands-on tests. It ships with a 1-million-token context window at introductory pricing of $2 / $10 per 1M tokens through August 31, 2026 (then $3 / $15), and is the new free and Pro default on claude.ai, so most writers get it at no cost; note its updated tokenizer maps the same text to roughly 1.0-1.35x more tokens, which narrows the real-world price gap. GPT-5.5 stays the safer default for fact-anchored writing like reports and briefs, the older Claude Sonnet 4.6 remains a cheaper legacy option at $3 / $15, Gemini 3.5 Flash is the price-performance pick for bulk content, and Claude Opus 4.8 is the call for long-form revision where you want the model to push back on weak arguments.
Model | Best For | Strength | Weakness | Price (per 1M tokens) |
|---|---|---|---|---|
Claude Sonnet 5 | Style + GDPval-AA v2 knowledge work | 1,618 GDPval-AA v2, edges Opus 4.8 (1,615); behind Fable 5, 1M context | New tokenizer inflates token counts ~1.0-1.35x | $2 / $10 intro (then $3 / $15) |
GPT-5.5 | Business writing, factual reports | Improved factual reliability vs GPT-5.4 (OpenAI eval) | Style less expressive than Sonnet 5 | $5 / $30 |
Gemini 3.5 Flash | Bulk content, drafts at scale | Near-Sonnet quality, 40% cheaper than Pro | Weaker on hardest reasoning | $1.50 / $9.00 |
Claude Opus 4.8 | Long-form, high-stakes copy | Best editor for argument structure | Most expensive option here | $5 / $25 |
Claude Sonnet 4.6 | Budget Claude writing | Prior Sonnet, 1,395 GDPval-AA v2 | Superseded by Sonnet 5 | $3 / $15 |
Grok 4.3 | Casual, opinionated, X-style | Native X grounding, fewer guardrails | Not the natural pick for formal copy | $1.25 / $2.50 |
Runner-up and alternatives: Gemini 3.5 Flash is the runner-up for sheer volume at near-Sonnet quality, and GPT-5.5 is the runner-up for factual accuracy. Claude Opus 4.8 is the splurge pick for long-form. Grok 4.3 is the niche pick when you want X-style voice or live web context inside the draft.
What changed this month: Claude Sonnet 5 (June 30) is the headline writing launch, scoring 1,618 on GDPval-AA v2 to edge Opus 4.8 (1,615) as the first Sonnet-class model to top the concurrent Opus flagship, both behind Fable 5 (1,783), and it is now the free and Pro default on claude.ai. GPT-5.5 stays the pick for fact-anchored business writing, and Gemini 3.5 Flash remains the bulk-content value pick.
The best AI for everyday chat and daily assistant work is GPT-5.6, ChatGPT’s new default model as of July 9, 2026, with GPT-5.5 Instant as the proven fallback, Claude Opus 4.8 as the alternative when you want a more thoughtful tone, and Gemini 3.5 Flash as the budget alternative inside the free Gemini app. GPT-5.6 replaced GPT-5.5 as ChatGPT’s default when it reached general availability, and it is more capable across coding, biology, and cybersecurity; most ChatGPT users get the balanced Terra tier, which OpenAI says matches GPT-5.5 at roughly half the cost. It is available inside ChatGPT (free with limits, Plus at $20/month, Pro at $100/month for roughly 5x Plus usage or $200/month for roughly 20x Plus usage), through the API (Luna $1 / $6, Terra $2.50 / $15, Sol $5 / $30 per 1M tokens), and, alongside GPT-5.5, bundled inside Fello AI with Claude, Gemini, Grok, and DeepSeek. One caveat: OpenAI’s system card and the evaluator METR flagged elevated “scheming” behaviour in the Sol tier, so GPT-5.5 Instant, with its documented 52.5% drop in hallucinated claims over GPT-5.3 Instant, stays the safer pick for high-stakes factual work until independent GPT-5.6 numbers land.
Claude Opus 4.8 is the better pick when you want a model that pushes back on weak prompts and reasons more carefully through ambiguous questions; Gemini 3.5 Flash is the better pick when you are running everything through the free Gemini app or care about speed.
Model | Best For | Strength | Weakness | Price |
|---|---|---|---|---|
GPT-5.6 | Everyday chat, ChatGPT’s new default | More capable across coding, bio, cyber; Terra matches GPT-5.5 at ~half cost | Scheming flagged by METR; independent factuality benchmarks pending | Free / $20/mo Plus; API $1 / $6 to $5 / $30 |
GPT-5.5 Instant | Proven factual daily assistant | 52.5% fewer hallucinated claims vs 5.3 Instant | Less expressive than Claude Sonnet 5 | $20/mo Plus; gpt-5.5 API at $5 / $30 |
Claude Opus 4.8 | Thoughtful, nuanced answers | Strong reasoning, pushes back well | $25 output API is the priciest here | $20/mo Pro, $5 / $25 API |
Gemini 3.5 Flash | Fast, free, multimodal | Free in Gemini app, 1M context | Weaker on hardest reasoning | Free / $1.50 / $9.00 API |
Grok 4.5 | Live news, X integration | Real-time X & web grounding, Opus-class reasoning | Higher hallucination rate; no EU yet | $30/mo SuperGrok |
Fello AI | All five models, one app | ChatGPT + Claude + Gemini + Grok + DeepSeek | Routed via app, not direct | $9.99/mo |
Runner-up and alternatives: GPT-5.5 Instant is the runner-up as the proven-factuality default, Claude Opus 4.8 is the runner-up for thoughtful daily use, Gemini 3.5 Flash is the runner-up for fast/free, and Grok 4.5 is the niche pick for live-news heavy days. Fello AI is the natural pick if you want all five top models in one Mac/iOS app for $9.99/month instead of juggling subscriptions.
What changed this month: GPT-5.6 reached general availability on July 9 and became ChatGPT’s new default, ending the two-week gated preview that began June 26. Terra is the everyday tier OpenAI says matches GPT-5.5 at roughly half the cost ($2.50 / $15 API), Luna is the fast, cheapest tier ($1 / $6), and Sol is the flagship. We rank GPT-5.6 as the chat pick because it is now the model ChatGPT serves by default, but we keep GPT-5.5 Instant one notch back as the proven-factuality fallback given the METR scheming flag on Sol. Claude Opus 4.8 holds the #1 spot on the Artificial Analysis Intelligence Index at 61, and on the Claude side the free and Pro default is Claude Sonnet 5 (June 30), a cheaper near-Opus model at $2 / $10 introductory pricing.
The best AI for image generation is ChatGPT Images 2.0, with Google Nano Banana Pro (Gemini 3 Pro Image) as the alternative for photorealism, Reve 2.0 as the layout-and-typography alternative, and Midjourney v8 as the alternative for stylized art. ChatGPT Images 2.0 (April 21, 2026) leads on text rendering, multilingual scripts, and infographic-style output, which makes it the natural pick when your image needs to contain words. Google’s Nano Banana Pro (Gemini 3 Pro Image, with the lower-cost Nano Banana 2 / Gemini 3.1 Flash Image as its sibling) is the natural pick for photoreal portraits and product shots, priced around $0.134 per 1K/2K image and $0.24 per 4K image. Reve 2.0 (June 3) jumped to #2 on the Arena text-to-image leaderboard with native 4K output and editing that preserves typography. Midjourney v8 stays the niche choice for distinctive style.
Model | Best For | Strength | Weakness | Price |
|---|---|---|---|---|
ChatGPT Images 2.0 | Images with readable text | Best multilingual text rendering | Less photoreal than Nano Banana | Included in ChatGPT Plus |
Nano Banana Pro (Gemini 3 Pro Image) | Photoreal portraits, products | Photorealism, ~$0.134 per 1K/2K image | Style less distinctive | Gemini app / AI Studio |
Reve 2.0 | Layout, typography, native 4K | #2 Arena, 16MP output, layout editing | New, smaller ecosystem | Free / from $7.99/mo |
Midjourney v8 | Stylized art, illustration | Aesthetic baseline most artists like | Weaker on text in image | $10-$120/mo |
Seedream 5.0 Pro | Multilingual text + region-precise editing | 10+ languages incl. Arabic RTL, lasso/layer editing | New, no independent benchmarks; copyright cloud | BytePlus / Magnific (1.5K) |
Grok Imagine | NSFW / Spicy Mode | Most permissive guardrails | Smallest model behind | $30/mo SuperGrok |
MAI-Image-2.5 | Microsoft ecosystem | #3 text-to-image leaderboard, native in Copilot | Just launched, US-first | Included in Copilot |
Runner-up and alternatives: Nano Banana Pro is the runner-up overall and the leader for photoreal work; Reve 2.0 is the runner-up for layout and typography; Midjourney v8 is the niche pick for art-direction-heavy use. Grok Imagine is the only major model that allows Spicy Mode adult content.
What changed this month: ByteDance launched Seedream 5.0 Pro on July 8, a multilingual image model built for text-heavy layouts (10+ languages including right-to-left Arabic) and region-precise editing (click, lasso, recolor, swap materials, separate layers). It challenges ChatGPT Images 2.0 on text rendering, but there are no independent benchmarks yet, access is via BytePlus and Magnific rather than a consumer app, and it carries the same Hollywood copyright scrutiny as ByteDance’s Seedance line, so ChatGPT Images 2.0 keeps the top spot for now. Reve 2.0 (June 3) still holds #2 on the Arena text-to-image leaderboard with native 4K rendering and layout-based editing, and Microsoft’s MAI-Image-2.5 (June 2) sits at #3, native in Copilot.
The best AI for video generation is Google Veo 3.1, with Kling 3.5 as the alternative for fast iteration and Runway Gen-4 as the alternative for cinematic motion control. OpenAI retired the Sora 2 consumer web and app experience on April 26, 2026 (the Sora 2 API remains available to developers until September 24, 2026), so OpenAI no longer ranks in this consumer category. Veo 3.1 is available inside the Gemini app, Google AI Studio, and via Vertex AI, with native audio generation, 1080p output, and the strongest physics consistency in the current lineup. Kling 3.5 stays the speed pick at lower cost; Runway Gen-4 is the choice when you need precise camera control. Pika 2.0 and Luma Ray 3 remain credible alternatives for shorter clips.
Model | Best For | Strength | Weakness | Price |
|---|---|---|---|---|
Google Veo 3.1 | Highest-fidelity AI video + audio | 1080p, native audio, physics consistency | Compute-heavy, slower | Gemini AI Pro / Ultra |
Kling 3.5 | Fast iteration | Quick turnaround, strong motion | Less stable on long shots | From $10/mo |
Runway Gen-4 | Cinematic control | Best-in-class camera/motion control | Pricing premium | Free / $12 mo billed annually, or $15 monthly |
Pika 2.0 | Short clips, social | Cheap, fast, easy UX | Lower max resolution | From $10/mo |
Luma Ray 3 | Photoreal scenes | Strong realism for landscapes | Smaller community | Free / from $9.99/mo |
Runner-up and alternatives: Kling 3.5 is the runner-up overall and the cost-conscious pick; Runway Gen-4 is the runner-up for filmmakers and ad teams. Sora 2’s consumer app is retired; only the developer API remains, through September 24, 2026.
What changed this month: No major video launches in July 2026, so Veo 3.1 stays uncontested at the top of the still-supported video models. Google is widely expected to refresh Veo at its next AI event; we will update this section when that happens.
The best AI for coding is Claude Fable 5, which returned on July 1 after the US government lifted the June 12 export-control order that had pulled it offline. Anthropic’s Mythos-class flagship retakes the coding crown at 80.3% on SWE-Bench Pro, the highest score of any model you can use, and is purpose-built for long-horizon autonomous runs at $10 / $50 per 1M tokens. Claude Opus 4.8 is the everyday-value pick right behind it, holding Anthropic’s top SWE-bench Verified score, remaining the favourite inside Claude Code and Cursor, and costing half as much at $5 / $25. The cheapest serious contender is xAI’s Grok 4.5 (July 8), a Cursor-trained model at just $2 / $6 per 1M tokens that Artificial Analysis now ranks #4 overall at Intelligence Index 54. LongCat-2.0, and Microsoft’s MAI-Code-1-Flash are the open-weight and budget alternatives.
Gemini 3.5 Flash (May 19) hit 76.2% on Terminal-Bench 2.1 and 83.6% on MCP Atlas at $1.50 / $9.00 per 1M tokens, making it the strongest price-performance option for agent workflows. On the open-weight side, MiniMax M3 (June 1) posts 59% on SWE-Bench Pro and 66% on Terminal-Bench 2.1 at roughly $0.60 per million input tokens, and Meituan’s new LongCat-2.0 (June 29, MIT) posts 59.5% on SWE-Bench Pro and 70.8 on Terminal-Bench, both edging GPT-5.5. Microsoft’s MAI-Code-1-Flash (June 2) beats Claude Haiku 4.5 on SWE-Bench Verified (71.6 vs 66.6) while using up to 60% fewer tokens, rolling out inside VS Code and the GitHub Copilot CLI. If you want to self-host, Kimi K2.6 (Modified MIT), GLM-5.2 (MIT, 1M context, built for long autonomous runs), and LongCat-2.0 are the strongest open-weight coders with clear commercial licenses.
Model | Best For | Strength | Weakness | Price (per 1M tokens) |
|---|---|---|---|---|
Claude Fable 5 | Best coding overall, long-horizon agentic | 80.3% SWE-Bench Pro, Mythos-class | Priciest; built for hours-long runs | $10 / $50 |
Claude Opus 4.8 | Everyday-value agentic coding | Anthropic-leading SWE-bench, adaptive thinking | Half the price, below Fable 5 on SWE-Bench Pro | $5 / $25 |
Grok 4.5 | Cheap value coder | 83.3% Terminal-Bench 2.1, II 54 (#4 overall), ~4x token efficiency | 64.7% SWE-Bench Pro (below Opus); higher hallucination rate; no EU yet | $2 / $6 |
Muse Spark 1.1 | Cheap agentic tool-use | 88.1 MCP Atlas, 1M context, self-managed agents | Trails Opus/GPT-5.5 on pure coding; US-only preview | $1.25 / $4.25 |
GPT-5.6 Sol | OpenAI flagship, agentic coding | 88.8% Terminal-Bench 2.1 (91.9% ultra) | SWE-bench not published; eval-gaming flagged by METR | $5 / $30 |
GPT-5.5 | Frontier proprietary alternative | 58.6% SWE-Bench Pro, 82.7% Terminal-Bench 2.0 | Less agent-tuned than Claude | $5 / $30 |
Gemini 3.5 Flash | Agent coding at scale | 76.2% Terminal-Bench, 83.6% MCP Atlas | Weaker on hardest reasoning | $1.50 / $9.00 |
LongCat-2.0 | Open-weight frontier coder | 59.5% SWE-Bench Pro, MIT, 1M context | New, self-host/provider only | Open weights (MIT) |
MiniMax M3 | Cheap frontier-class, self-host | 59% SWE-Bench Pro, 1M context, multimodal | Weights/license still rolling out | ~$0.60 input |
DeepSeek V4-Flash | Cheap open-weight coding | MIT, 1M context, II 47 | Below V4-Pro on hardest tasks | $0.14 / $0.28 |
Runner-up and alternatives: GPT-5.5 is the proprietary runner-up; Gemini 3.5 Flash is the runner-up for price-performance; Qwen 3.7 Max is the runner-up for mid-tier value; MiniMax M3, LongCat-2.0, and DeepSeek V4 are the runners-up for open-weight self-hosters. Inside IDEs, Cursor + Claude Opus 4.8 is the most popular pairing and Claude Code is the natural pick if you live in the terminal.
What changed this month: Claude Fable 5 returned on July 1 and retakes the best-coding pick at 80.3% SWE-Bench Pro, the highest of any usable model. xAI shipped Grok 4.5 publicly on July 8, a Cursor-trained coding model at $2 / $6; Artificial Analysis now ranks it #4 overall at Intelligence Index 54, matching the field on Terminal-Bench 2.1 (83.3%) but landing at 64.7% on SWE-Bench Pro, below Opus 4.8. Meta launched Muse Spark 1.1 (July 9), its first paid model, at $1.25 / $4.25, an agentic coder that tops tool-use benchmarks (88.1 MCP Atlas) but trails on pure SWE-Bench coding. Meituan open-sourced LongCat-2.0 (June 29), a 1.6T MIT model at 59.5% SWE-Bench Pro. OpenAI’s GPT-5.6 family reached general availability on July 9; its flagship Sol scores 88.8% on Terminal-Bench 2.1 (91.9% in ultra mode), just ahead of GPT-5.5’s 88.0%, though OpenAI withheld its SWE-bench Verified score and METR flagged Sol for gaming a coding eval. Open-weight coding now has a deep bench: MiniMax M3 (II 55), LongCat-2.0, Kimi K2.6, and GLM-5.2 all sit at or above 58-59% on SWE-Bench Pro under permissive or commercial licenses. Anthropic’s new Claude Sonnet 5 (June 30) posts 63.2% SWE-Bench Pro at $2 / $10 introductory pricing, a cheaper near-Opus Claude coder.
The best AI for creative writing, brainstorming, and unfiltered ideation is Grok 4.5, xAI’s new public flagship as of July 8, with Claude Opus 4.8 as the alternative for structured creative work and Gemini 3.1 Pro as the alternative for multimodal creative tasks. Grok 4.5 is now the default model in the Grok app for SuperGrok and X Premium+ subscribers, and it keeps the Grok line’s permissive guardrails and strongest-in-class native X integration while adding Opus-class reasoning, which makes it the natural pick for opinionated, on-trend, real-time creative work. Grok 4.3 stays the cheaper fallback on the free and lower tiers. Claude Opus 4.8 is the better pick when you want a model that holds a long creative thread, edits its own drafts, and engages with the substance of your work. Gemini 3.1 Pro is the better pick when your creative project mixes text with images, video, and live web context.
Model | Best For | Strength | Weakness | Price |
|---|---|---|---|---|
Grok 4.5 | Unfiltered, opinionated, on-trend | Opus-class reasoning + fewest guardrails, X integration | Newer; creative prose still maturing | $30/mo SuperGrok |
Claude Opus 4.8 | Long-form structured creativity | Holds long threads, self-edits | Most cautious of the four | $20/mo Pro, $5 / $25 API |
Gemini 3.1 Pro | Multimodal creative | Strong text + image + video chain | Quotas inside Gemini app | Free / $2.00-$4.00 API in |
ChatGPT-5.5 | Mainstream creative writing | Best at hitting briefs | Heavier guardrails | $20/mo Plus, $5 / $30 API |
Grok Imagine (Spicy Mode) | NSFW / adult creative | Most permissive image generation | Niche use case | $30/mo SuperGrok |
Runner-up and alternatives: Claude Opus 4.8 is the runner-up overall and the right pick for projects that need to hold together across many turns. Gemini 3.1 Pro is the multimodal runner-up. For adult creative work, Grok Imagine Spicy Mode is the only frontier-grade option.
What changed this month: Grok 4.5 went public on July 8 and is now the default in the Grok app for SuperGrok and X Premium+ subscribers, so it takes over as our creativity pick from Grok 4.3, adding Opus-class reasoning on top of the Grok line’s permissive guardrails and native X integration. Grok 4.3 stays the cheaper fallback on the free and lower tiers.
The best AI for accuracy and research is Gemini 3.1 Pro, with Qwen 3.7 Max as the value alternative and GPT-5.5 Pro as the alternative for hallucination-sensitive work. Gemini 3.1 Pro leads the hardest pure-reasoning tests at 94.3% on GPQA Diamond, 44.4% on Humanity’s Last Exam, and 77.1% on ARC-AGI-2, with native Google Search grounding for live factual answers. Qwen 3.7 Max (May 20) entered the top tier at 92.4 on GPQA Diamond, tied with Claude Opus 4.8, at half the API cost.
GPT-5.5 Pro (April 23) carries GPT-5.5’s factual-reliability gains over GPT-5.4 (claims 23% more likely to be factually correct on OpenAI’s flagged-conversation set), which makes it the right pick when factual reliability matters more than raw benchmark depth. Gemini 3.5 Flash (May 19) outscores Gemini 3.1 Pro on coding and agent benchmarks but trails Pro on these accuracy tests (HLE 40.2% vs 44.4%, ARC-AGI-2 72.1% vs 77.1%), so Pro stays the accuracy pick.
Model | Best For | Key Benchmark | Weakness | Price |
|---|---|---|---|---|
Gemini 3.1 Pro | Hardest reasoning + research | 94.3% GPQA, 44.4% HLE, 77.1% ARC-AGI-2 | API quotas in app | $2.00-$4.00 / $12.00-$18.00 (tiered) |
Qwen 3.7 Max | Frontier accuracy at value pricing | 92.4 GPQA Diamond | API-only, no chat front-end | $1.25 / $3.75 promo; $2.50 / $7.50 list |
GPT-5.5 Pro | Hallucination-sensitive work | Improved factual reliability vs GPT-5.4 (OpenAI eval) | Pricier API tier | $100/mo ChatGPT Pro |
Claude Opus 4.8 | Long-form factual writing | #1 Intelligence Index (61) | Slower on hardest math | $5 / $25 |
Grok 4.3 | Live web facts | Native real-time grounding | Smaller benchmark coverage | $30/mo SuperGrok |
Runner-up and alternatives: Qwen 3.7 Max is the runner-up and the value pick at the frontier. GPT-5.5 Pro is the runner-up for hallucination-sensitive work. Claude Opus 4.8 is the runner-up for long-form factual writing.
What changed this month: No new accuracy leaders shipped, so Gemini 3.1 Pro holds the top of the category. The one to watch is Gemini 3.5 Pro, now cleared for a July general-availability launch after slipping from June; its specs are not yet public, but it could reset this ranking the moment it reaches general availability.
The best AI for hard problem-solving is GPT-5.6 Sol, OpenAI’s new flagship for abstract math and science, and Qwen 3.7 Max for competition math, with Claude Opus 4.8 as the alternative for long agentic reasoning chains. GPT-5.6 Sol (July 9) is purpose-tuned for the hardest STEM reasoning and is the natural pick when you need step-by-step working on tough math and physics; OpenAI has not yet published Sol’s FrontierMath number, so the verified OpenAI benchmark remains GPT-5.5 Pro’s 39.6% on FrontierMath Tier 4 (nearly double Claude Opus 4.8’s 22.9%), and we will slot in Sol’s score the moment it goes public. Qwen 3.7 Max (May 20) hit 97.1 on HMMT 2026 February, the highest score in its comparison group, and 44.5 on Apex, which makes it the right pick for competition-style problem-solving at a fraction of the cost of ChatGPT Pro.
Claude Opus 4.7 (April 16) introduced task budgets, a primitive for guiding agentic token spend on long chains; Claude Opus 4.8 (May 28) instead uses adaptive thinking controlled by an effort parameter, and does not support extended-thinking budgets. Gemini 3.5 Flash trades raw reasoning depth for speed and price; for the hardest problems, Gemini 3.1 Pro and the Thinking variants still lead.
Model | Best For | Key Benchmark | Weakness | Price |
|---|---|---|---|---|
GPT-5.6 Sol | Abstract math, science (new flagship) | STEM-tuned successor to 5.5 Pro; FrontierMath not yet published | Scheming flagged by METR; numbers pending | $100/mo ChatGPT Pro; API $5 / $30 |
GPT-5.5 Pro | Verified FrontierMath leader | 39.6% FrontierMath Tier 4 | Superseded by Sol as flagship | $100/mo ChatGPT Pro |
Qwen 3.7 Max | Competition math | 97.1 HMMT 2026 Feb, 44.5 Apex | API-only | $1.25 / $3.75 promo; $2.50 / $7.50 list |
Claude Opus 4.8 | Long agentic reasoning | Adaptive thinking, effort control, #1 Intelligence Index | Slower on math | $5 / $25 |
Gemini 3.1 Pro | Multimodal reasoning + research | 94.3 GPQA, 77.1 ARC-AGI-2 | API quotas | $2.00-$4.00 / $12.00-$18.00 (tiered) |
DeepSeek V4-Flash | Open-weight problem solving | MIT, 1M context, II 47 | Below V4-Pro on hardest | $0.14 / $0.28 |
Runner-up and alternatives: Claude Opus 4.8 is the runner-up overall and the natural pick for agentic, long-chain problem-solving. Gemini 3.1 Pro is the multimodal runner-up. DeepSeek V4-Flash is the open-weight runner-up.
What changed this month: GPT-5.6 Sol reached general availability on July 9 and takes over as OpenAI’s flagship problem-solver as the STEM-tuned successor to GPT-5.5 Pro. OpenAI has not yet published Sol’s FrontierMath or AIME numbers, so the verified OpenAI benchmark on the page stays GPT-5.5 Pro’s 39.6% on FrontierMath Tier 4 until Sol’s score is public; Qwen 3.7 Max still leads competition math at 97.1 HMMT 2026 February. The open-weight MiniMax M3 and LongCat-2.0 remain strong cheaper options for agentic reasoning chains.
The best AI agent right now is Gemini Spark for 24/7 cloud-resident work and Claude Cowork for desktop-resident work, with ChatGPT Codex as the alternative for coding agents and OpenAI Operator-class browser agents as the alternative for web tasks. AI agents are the fastest-moving category of 2026: each top vendor now ships an agent product, and the practical choice is between agents that live in the cloud (run while your laptop is closed) and agents that live on your desktop (drive your apps directly).
Gemini Spark launched at Google I/O on May 19, 2026 and is the first 24/7 cloud agent. Claude Cowork launched in general availability on April 9, 2026 and runs as a desktop agent that drives your local apps. ChatGPT Codex Mobile (May 14) is the pick for coding-agent work, now usable from iOS and Android. Read the full Gemini Spark vs Claude Cowork comparison.
Agent | Best For | Where It Runs | Strength | Price |
|---|---|---|---|---|
Gemini Spark | 24/7 cloud tasks, Workspace workflows | Google Cloud VM (always-on) | First true 24/7 agent, deep Workspace integration | $100/mo Google AI Ultra |
Claude Cowork | Desktop, app-driving, design + code | Your Mac/Windows desktop | Drives local apps, sees your screen | $20/mo Claude Pro |
ChatGPT Codex Mobile | Coding agent on phone | OpenAI cloud + iOS/Android | Approve diffs and redirect work from phone | Included in ChatGPT plans |
Grok Agentic (Grok 4.3) | Real-time research, X scraping | xAI cloud | Native X integration | $30/mo SuperGrok |
OpenAI Operator-class | Browser tasks, web forms | OpenAI cloud + your browser | Web automation | ChatGPT Pro |
Runner-up and alternatives: Claude Cowork is the runner-up overall and the natural pick when you want the agent on your machine driving your apps. ChatGPT Codex Mobile is the runner-up for coding agents. Grok Agentic is the niche pick for real-time research.
What changed this month: No new consumer agents shipped, so the Gemini Spark (cloud) vs Claude Cowork (desktop) choice still drives most agent decisions for individual users. With Claude Fable 5 back online, the strongest model you can run inside an agent system for long-horizon autonomous work is available again. For teams building their own agents, Meta’s new Muse Spark 1.1 (July 9) is a cheap agent-native option at $1.25 / $4.25, with built-in primary-agent and subagent orchestration and an API that speaks both the OpenAI and Anthropic formats, and the open models LongCat-2.0, NVIDIA Nemotron 3 Ultra, and MiniMax M3 all ship strong agentic benchmarks.
Here is the July 2026 pricing comparison for every leading AI model, in API cost per 1 million tokens and the consumer-subscription price for the same model. Free tiers exist for ChatGPT, Gemini, Claude, Grok, and DeepSeek. This week’s launches undercut the field on price: Meta’s Muse Spark 1.1 lists at $1.25 / $4.25 and Grok 4.5 at $2 / $6, and on a price-per-intelligence basis Artificial Analysis puts a Grok 4.5 index task at about $0.31, five times cheaper than Claude Sonnet 5. Among closed models Gemini 3.5 Flash at $1.50 / $9.00 stays the cheapest frontier all-rounder; the cheapest open-weight frontier coder is MiniMax M3 at around $0.60 per million input tokens, and the cheapest open-weight model with a 1M context is DeepSeek V4-Flash at $0.14 / $0.28. For a deeper breakdown by tier, see our full AI Pricing Comparison Guide hub.
Model | Input (per 1M) | Output (per 1M) | Context Window | Free access? |
|---|---|---|---|---|
GPT-5.5 | $5.00 | $30.00 | 1M (400K in Codex) | ChatGPT Free; API paid |
GPT-5.5 Pro | $30.00 | $180.00 | 1M | ChatGPT Pro from $100/mo ($200 higher-usage tier) |
GPT-5.6 Sol | $5.00 | $30.00 | not published | Live in ChatGPT, Codex & API (July 9) |
GPT-5.6 Terra | $2.50 | $15.00 | not published | Live in ChatGPT, Codex & API (July 9) |
GPT-5.6 Luna | $1.00 | $6.00 | not published | Live in ChatGPT, Codex & API (July 9) |
Claude Opus 4.8 | $5.00 | $25.00 | 1M | No (Pro/Max/API) |
Claude Fable 5 | $10.00 | $50.00 | 1M | Pro/Max/Team (up to 50% weekly limits to Jul 7, then credits); API |
Claude Sonnet 5 | $2.00 intro / $3.00 list | $10.00 intro / $15.00 list | 1M | Claude Free & Pro default; API paid |
Claude Sonnet 4.6 | $3.00 | $15.00 | 1M | API paid (superseded by Sonnet 5) |
Gemini 3.1 Pro | $2.00 (≤200K) / $4.00 (>200K) | $12.00 (≤200K) / $18.00 (>200K) | 1M | Limited Gemini app; API paid |
Gemini 3.5 Flash | $1.50 | $9.00 | 1M | Gemini app/AI Studio; free API tier + paid API |
Qwen 3.7 Max | $1.25 promo / $2.50 list | $3.75 promo / $7.50 list | 1M | API only |
MiniMax M3 | ~$0.60 | ~$2.40 (≤512K) | 1M | Open weights; hosting costs apply |
LongCat-2.0 | Provider-dependent | Provider-dependent | 1M | Open weights (MIT); hosting costs apply |
NVIDIA Nemotron 3 Ultra | Provider-dependent | Provider-dependent | 1M | Open weights (OpenMDW); hosting costs apply |
Qwen 3.5 (open-weight) | Self-host / Together | Self-host / Together | 1M | Open weights; hosting costs apply |
Nex-N2-Pro | Self-host / providers | Self-host / providers | 1M | Open weights (Apache 2.0); hosting costs apply |
Rio 3.5 Open 397B | Self-host / providers | Self-host / providers | 1M | Open weights (MIT); hosting costs apply |
Grok 4.3 | $1.25 | $2.50 | 1M | Free consumer plan; API paid |
Grok 4.5 | $2.00 | $6.00 | 500K | Grok Build / Cursor / xAI console; no EU yet |
Muse Spark 1.1 | $1.25 | $4.25 | 1M | Meta Model API ($20 free credits, US preview); free in Meta AI Thinking mode |
DeepSeek V4-Pro | $0.435 ($0.0036 cache-hit) | $0.87 | 1M | DeepSeek Chat free; API paid |
DeepSeek V4-Flash | $0.14 | $0.28 | 1M | DeepSeek Chat free; API paid |
Kimi K2.7 Code | Provider-dependent | Provider-dependent | 256K | Open weights; hosting costs apply |
GLM-5.2 | Provider-dependent | Provider-dependent | 1M | Open weights; hosting costs apply |
ERNIE 5.1 | China-region pricing | China-region pricing | 256K | Baidu free tier |
Gemini Spark (agent) | Not API-priced | Not API-priced | 1M (Gemini base) | Google AI Ultra $100 or $200/mo |
Fello AI (aggregator) | Routed via app | Routed via app | Model-dependent | $9.99/mo |
The GPT-5.5 and GPT-5.5 Pro rates above are short-context prices; prompts over 272K input tokens bill at the long-context rate of $10 / $45 and $60 / $270 per 1M tokens respectively.
If you want access to multiple AI models without managing separate subscriptions, Fello AI provides GPT, Claude, Gemini, Grok, Perplexity, and more in a single app for Mac, iPhone, and iPad, starting at $9.99/month with a free tier available. Models are updated regularly so you always have access to the latest.
Claude hit #1 on the App Store in early 2026, pushing ChatGPT out of the top spot for the first time. The catalyst was Anthropic publicly refusing the Pentagon’s demand to deploy its models for autonomous weapons and mass surveillance, after which the government labelled Anthropic a “supply chain risk.”
Update, July 10, 2026: Both chatbots just moved to new flagship models. ChatGPT now runs GPT-5.6 (the Sol, Terra, and Luna tiers), which began its broad public rollout on July 9, 2026, and Grok is powered by Grok 4.5, xAI’s coding-focused release from July 8, 2026. The pricing, benchmarks, and
Update, July 10, 2026: ChatGPT moved to a new flagship. GPT-5.6 began its broad public rollout on July 9, 2026 as a three-tier family, Sol (flagship), Terra (balanced), and Luna (fast and cheapest). On the Google side, Gemini 3.1 Pro remains the paid flagship on Google AI Pro and Gemini
Choosing the right AI tool in 2026 feels like trying to hit a moving target. New models arrive every few weeks, and what worked best in January might already be outdated today. This guide cuts through the hype to show you exactly which tools are winning right now based on
TL;DR: In January 2026, there isn’t one “best” AI for everything. On LMArena’s Text leaderboard, Gemini 3 Pro leads user-preference rankings, while the updated Artificial Analysis Intelligence Index v4.0 reports GPT-5.2 (with extended reasoning) as the top overall benchmark performer. Choose based on your task: Gemini for daily assistance, Claude
Update, July 2026: both models here have since been succeeded. Google has shipped Nano Banana 2 and Nano Banana 2 Pro, and OpenAI’s GPT Image line now powers ChatGPT Images 2.0. The head-to-head below is our original December 2025 test of GPT-Image-1.5 vs Nano Banana Pro, with the hands-on images
The best AI for students is GPT-5.5 Free inside ChatGPT for general coursework and Gemini 3.5 Flash Free inside the Gemini app for STEM and multimodal study, with Qwen 3.7 Max as the API alternative for harder problem sets and Claude Opus 4.8 as the alternative for essay editing.
Most students don’t need to pay: the free ChatGPT tier now defaults to GPT-5.6 (with GPT-5.5 still available), Gemini 3.5 Flash is in the free Gemini app and AI Studio, Claude Sonnet 5 is the new free Claude default, and DeepSeek V4 is free on DeepSeek’s chat site. For step-by-step working on the hardest math, GPT-5.6 Sol is OpenAI’s new flagship (its FrontierMath score is not yet published, with GPT-5.5 Pro’s verified 39.6% on FrontierMath Tier 4 the current mark), though both are paid-only; Qwen 3.7 Max is the value alternative at 97.1 HMMT 2026 February with API pricing at $1.25 / $3.75 on its current 50% promo ($2.50 / $7.50 list).
Task | Best Model | Why | Free? | Alternative |
|---|---|---|---|---|
Essays & coursework | GPT-5.5 | Free in ChatGPT, improved factual reliability vs 5.4 | Yes | Claude Sonnet 5 (free Claude) |
STEM problem-solving | GPT-5.6 Sol / Qwen 3.7 Max | New STEM flagship (5.5 Pro: 39.6% FrontierMath) / 97.1 HMMT 2026 Feb | Pro paid / Qwen API paid | Gemini 3.5 Flash (free) |
Research & accuracy | Gemini 3.1 Pro | Native Google Search grounding | Yes (Gemini app) | Claude Opus 4.8 |
Writing editing | Claude Sonnet 5 | Best instruction-following, edges Opus 4.8 on GDPval-AA v2 | Yes (Claude free) | GPT-5.5 |
Multimodal study (PDFs, slides, images) | Gemini 3.5 Flash | 1M context, free in Gemini app | Yes | NotebookLM (Google) |
Runner-up and alternatives: Claude Sonnet 5 (free) is the runner-up for essay writing and editing. Gemini 3.5 Flash (free) is the runner-up for multimodal study and PDF ingestion. DeepSeek V4 is the runner-up for problem-solving on a strict zero-cost budget.
The best AI for professional work is GPT-5.6 (ChatGPT’s new default since July 9, with GPT-5.5 the proven fallback) for daily knowledge work, Claude Opus 4.8 for coding and high-stakes writing, and Gemini Spark for 24/7 agentic workflows. Most professionals get the most out of running two paid subscriptions (ChatGPT Plus at $20/month plus Claude Pro at $20/month, total $40/month), or consolidating with Fello AI at $9.99/month for all five top models in one Mac/iOS app. For agentic work that runs while you sleep, Gemini Spark on Google AI Ultra at $100/month is the only true 24/7 cloud agent.
Use Case | Best Model | Key Stat | Price | Alternative |
|---|---|---|---|---|
Daily knowledge work | GPT-5.6 | ChatGPT’s new default (July 9); GPT-5.5 the proven fallback | $20/mo ChatGPT Plus | Claude Opus 4.8 |
Coding (proprietary) | Claude Opus 4.8 | Anthropic-leading SWE-bench | $20/mo Claude Pro | GPT-5.5 |
Coding (cost-effective) | Qwen 3.7 Max | 80.4 SWE-Verified, 1M context | $1.25 / $3.75 promo; $2.50 / $7.50 list | MiniMax M3 |
Research & briefings | Gemini 3.1 Pro | 94.3% GPQA Diamond, Google grounding | Google AI Pro / Ultra | Claude Opus 4.8 |
Hard math, physics, finance modelling | GPT-5.6 Sol | OpenAI’s new STEM flagship (5.5 Pro verified at 39.6% FrontierMath) | $100/mo ChatGPT Pro | Qwen 3.7 Max |
Always-on agent workflows | Gemini Spark | First 24/7 cloud agent | $100/mo Google AI Ultra | Claude Cowork |
Live news, X-context creative | Grok 4.5 | Opus-class + native X grounding | $30/mo SuperGrok | Gemini 3.1 Pro |
All-in-one consolidation | Fello AI | ChatGPT + Claude + Gemini + Grok + DeepSeek | $9.99/mo | Pay each vendor separately |
Runner-up and alternatives: For most professional teams, Claude Opus 4.8 is the runner-up to GPT-5.5 for daily work and the leader for coding. Gemini 3.1 Pro is the runner-up for research-heavy roles, and Gemini Spark is the unique pick if you can put a cloud agent to work on long tasks.
The open-weight tier had its biggest shake-up of the year, and it is now genuinely crowded at the frontier. The newest arrival is Meituan’s LongCat-2.0 (June 29), a 1.6-trillion-parameter MoE under an MIT license that posts 59.5% on SWE-Bench Pro and 70.8 on Terminal-Bench, trained entirely on domestic Chinese chips. Ranked by Artificial Analysis Intelligence Index, the strongest measured open models are MiniMax M3 at 55, Kimi K2.6 and Xiaomi’s MiMo-V2.5-Pro at 54, DeepSeek V4-Pro at 52, and GLM-5.1 at 51.4, with NVIDIA Nemotron 3 Ultra at 48 as the most capable model under a fully permissive license.
Several lines refreshed in the last month: Moonshot released Kimi K2.7 Code (June 12), Zhipu released GLM-5.2 (June 13), and Meituan open-sourced LongCat-2.0 (June 29), whose independent Intelligence Index scores are still pending. Licensing now matters as much as raw score: MiniMax M3’s terms were unconfirmed at launch, while Kimi K2.7 (Modified MIT), DeepSeek V4 (MIT), GLM-5.2 (MIT), LongCat-2.0 (MIT), and Nemotron 3 Ultra (OpenMDW) all clearly allow commercial use. Nex-N2-Pro (Nex AGI, Apache 2.0, 397B parameters) posts the strongest open coding scores here at 80.8 on SWE-Bench Verified and 75.3 on Terminal-Bench 2.1, putting it alongside GPT-5.5 on agentic coding. Rio 3.5 Open 397B (IplanRIO, MIT, 17B active) is a Qwen 3.5-397B-A17B fine-tune that reportedly beats Qwen 3.7 Plus on four of five first-party benchmarks, though those results are not yet independently verified.
Model | Best For | Key Benchmark | Context / License | Where To Run |
|---|---|---|---|---|
LongCat-2.0 | Newest frontier open coder | 59.5% SWE-Bench Pro, 70.8 Terminal-Bench, 1.6T/~48B active | 1M / MIT | Hugging Face, GitHub, OpenRouter |
MiniMax M3 | Highest open Intelligence Index | II 55, 59% SWE-Bench Pro, multimodal | 1M / license TBD | Hugging Face, API ~$0.60/1M |
Nex-N2-Pro | Strongest open coding score | 80.8 SWE-Bench Verified, 75.3 Terminal-Bench 2.1, 397B/17B active | Qwen-based / Apache 2.0 | Hugging Face, providers, self-host |
Kimi K2.7 Code | Strongest commercially-licensed open coder | +21.8% on Kimi Code Bench v2 vs K2.6 (vendor); 1T/32B active | 256K / Modified MIT | Hugging Face, DeepInfra, providers |
DeepSeek V4-Pro | Agentic real-world work | II 52, 1.6T/49B active | 1M / MIT | DeepSeek API ($0.435/$0.87), local |
GLM-5.2 | Long-horizon agentic coding, 1M context | 744B/40B active, coding-first; independent benchmarks pending | 1M / MIT | Z.ai, Hugging Face, OpenRouter |
NVIDIA Nemotron 3 Ultra | Most capable permissive-license open | II 48, 65-70.4 SWE-Bench Verified, 550B/55B active | 1M / OpenMDW | OpenRouter, Hugging Face, AWS (8× B200 self-host) |
DeepSeek V4-Flash | Cheapest 1M-context open model | II 47, $0.14/$0.28 per 1M, 284B/13B active | 1M / MIT | DeepSeek API, local |
Qwen 3.5 (397B / 17B active) | Multimodal, fast decode | 88.4 GPQA, 91.3 AIME 2026, 83.6 LiveCodeBench v6 | 1M / open | Together, OpenRouter, local |
Qwen3.6-35B-A3B | Efficient open agentic coder (3B active) | 86.0 GPQA Diamond, 92.7 AIME 2026, 35B/3B active | 262K (→1M YaRN) / Apache 2.0 | Hugging Face, OpenRouter, local |
Qwen3.6-27B | Laptop-runnable dense coder | 87.8 GPQA Diamond, dense 27B, multimodal | 256K / Apache 2.0 | Local Mac/PC, Hugging Face, OpenRouter |
Rio 3.5 Open 397B | Qwen 3.5 fine-tune, multilingual reasoning | 70.8 Terminal-Bench 2.1 (first-party), beats Qwen 3.7 Plus on 4/5 | 397B / 17B active, MIT | Hugging Face, providers, self-host |
Qwen 3.5-9B | Laptop-runnable open-weight | 81.7 GPQA Diamond | Dense / open | Local Mac/PC with 16GB+ RAM |
Llama 4 Maverick | Meta-line flagship | 17B active / 400B total params | Llama 4 license | Meta cloud, Hugging Face, local |
NVIDIA Nemotron 3 Nano Omni | Edge / low-power | Multimodal, very small footprint | Compact / open | Local, NVIDIA tool |
Runner-up and alternatives: LongCat-2.0 and Kimi K2.7 Code are the newest picks when you need a clearly commercial license; MiniMax M3 still holds the highest measured Intelligence Index. DeepSeek V4-Pro and GLM-5.2 are the runners-up for agentic coding, DeepSeek V4-Flash is the cheapest way to get a 1M-context open model, and NVIDIA Nemotron 3 Nano Omni remains the natural pick for edge and on-device use. Xiaomi’s MiMo-V2.5-Pro (Intelligence Index 54) is also worth tracking as a fast-rising new entrant.
Every ranking on this page combines three inputs: public benchmarks (Artificial Analysis Intelligence Index, GPQA Diamond, ARC-AGI-2, Humanity’s Last Exam, SWE-bench Verified, GDPval-AA, FrontierMath, HMMT, Terminal-Bench, MCP Atlas, LM Arena), published API and subscription pricing from each vendor’s official pricing page, and hands-on use by the FelloAI editorial team running real prompts across the same task on every model. We re-fetch official pricing and benchmark sources before every monthly update.
Benchmarks are weighted to the use case: SWE-bench and Terminal-Bench drive coding, GPQA Diamond and ARC-AGI-2 drive accuracy, GDPval-AA (Artificial Analysis’s professional-deliverables benchmark) informs professional-task quality while writing style is judged primarily by hands-on testing, FrontierMath and HMMT drive problem-solving. We disclose when a benchmark is vendor-reported but not independently verified, and we strip any claim we cannot reproduce against a live source. When a model goes through a major upgrade between updates, we re-rank the category and add a “What changed this month” line at the bottom of the deep-dive.
It depends on the task. For daily chat and general assistance, GPT-5.6 became ChatGPT’s new default on July 9, with GPT-5.5 Instant (OpenAI reports 52.5% fewer hallucinated claims than GPT-5.3 Instant) the proven fallback. For coding, the newly returned Claude Fable 5 leads at 80.3% on SWE-Bench Pro, the highest of any usable model, with Claude Opus 4.8 (the #1 Intelligence Index model at 61) the everyday-value pick right behind it. For writing, the new Claude Sonnet 5 (June 30) edges Opus 4.8 on Artificial Analysis’s GDPval-AA v2 knowledge-work benchmark (1,618 vs 1,615, both behind Fable 5’s 1,783) and leads on writing style in hands-on tests. For accuracy and research, Gemini 3.1 Pro at 94.3% GPQA Diamond and 44.4% Humanity’s Last Exam. For hard math, GPT-5.6 Sol is OpenAI’s new STEM flagship (FrontierMath score pending; GPT-5.5 Pro verified at 39.6% on Tier 4), and Qwen 3.7 Max hit 97.1 HMMT 2026 February. For images, ChatGPT Images 2.0 leads on text rendering and Reve 2.0 is #2 on the Arena leaderboard. For agents, Gemini Spark is the first 24/7 cloud agent and Claude Cowork is the leader on desktop.
The month opens with Claude Fable 5 back online (July 1) after the US government lifted the June 12 export-control order that had pulled it offline. On July 8 and 9, four more launches landed: OpenAI’s GPT-5.6 family (Sol, Terra, Luna) reached general availability on July 9 and became ChatGPT’s new default; xAI took Grok 4.5 public on July 8 as a cheap Cursor-trained coding model at $2 / $6 per 1M tokens that Artificial Analysis ranks #4 overall (Intelligence Index 54); Meta shipped Muse Spark 1.1 on July 9 as its first paid model, a $1.25 / $4.25 agentic coder; and ByteDance released Seedream 5.0 Pro, a multilingual text-and-layout image model with region-precise editing. The rest of the headlines landed in the final week of June: OpenAI first previewed GPT-5.6 on June 26 behind a US-government access list; Meituan open-sourced LongCat-2.0, a 1.6-trillion-parameter MIT coding model trained entirely on Chinese chips, on June 29; and Anthropic made Claude Sonnet 5 its new default model on June 30. Gemini 3.5 Pro is now cleared for a July general-availability launch after slipping from June. These follow the late-May and June board of Claude Opus 4.8 (#1 at 61), Qwen 3.7 Max, Gemini 3.5 Flash, Gemini Spark, and the open-weight wave of MiniMax M3, NVIDIA Nemotron 3 Ultra, Kimi K2.7, and GLM-5.2.
Yes. Anthropic redeployed Claude Fable 5 on July 1, 2026 after the US government lifted the export-control restriction it had imposed on June 12. It is available again on the Claude API, Claude.ai, Claude Code, and Claude Cowork. For Pro, Max, Team, and select Enterprise plans it is included for up to 50% of weekly usage limits through July 7, after which it runs on usage credits; API pricing is $10 / $50 per million tokens. Fable 5 is a Mythos-class model built for long-horizon agentic work with a 1M-token context, and it reclaims the coding crown at 80.3% on SWE-Bench Pro.
Claude Sonnet 5 is Anthropic’s new default model, launched June 30, 2026 for Free and Pro users on claude.ai and live in the Claude API, Claude Code, Cursor, VS Code, and GitHub Copilot. It ships with a 1-million-token context window at introductory pricing of $2 / $10 per 1M tokens through August 31, 2026 (then $3 / $15). It takes the writing crown on style and instruction-following, scores 1,618 on GDPval-AA v2 to edge Opus 4.8 (1,615, both behind Fable 5’s 1,783), and closes much of the agentic gap to Opus 4.8 at 63.2% SWE-Bench Pro, while costing far less than GPT-5.5. The one caveat is an updated tokenizer that maps the same text to roughly 1.0-1.35x more tokens.
Yes, as of July 9, 2026. GPT-5.6 is OpenAI’s next-generation model family, and after a two-week gated preview that began June 26 behind a US-government safety review, it reached general availability on July 9. There are three tiers, least to most capable: Luna, the fast, cheapest tier ($1 / $6 per 1M tokens); Terra, a balanced everyday model OpenAI says matches GPT-5.5 at roughly half the cost ($2.50 / $15); and Sol, the flagship tuned for biology, chemistry, and cybersecurity ($5 / $30). It is now live across ChatGPT, Codex, and the API as OpenAI’s default model. On the limited benchmarks OpenAI published, Sol scores 88.8% on Terminal-Bench 2.1 (91.9% in ultra mode) versus GPT-5.5’s 88.0%; OpenAI withheld the usual SWE-bench, GPQA, and FrontierMath numbers, and the context window is still not officially published. One caveat: OpenAI’s system card and the external evaluator METR flagged elevated “scheming” behaviour in Sol, so treat it carefully for high-stakes factual work until independent results settle.
Yes. xAI released Grok 4.5 publicly on July 8, 2026, its first flagship since SpaceX absorbed the company and went public as SPCX. Elon Musk describes it as “an Opus-class model, but faster, more token-efficient and lower cost,” and says internally it is “roughly comparable to Opus 4.7, but much faster.” It is Cursor-trained and aimed at coding and agentic work, priced at $2 / $6 per 1M tokens with a 500K-token context window, well under Opus 4.8’s $5 / $25. It is live in Grok Build, Cursor on all plans, and the xAI console, with EU access expected mid-July. Independent scores are now in: Artificial Analysis ranks it #4 overall at Intelligence Index 54, matching the field on Terminal-Bench 2.1 (83.3%) but trailing on SWE-Bench Pro at 64.7% (below Opus 4.8’s 69.2%), so it is the value pick rather than the outright benchmark leader. Grok 4.5 also takes over as the pick in our creativity ranking, since it is the new default in the Grok app for SuperGrok and X Premium+ subscribers with the same permissive guardrails and native X integration.
MiniMax M3 (June 1) has the highest measured Intelligence Index of any open model at 55, though its license was unconfirmed at launch. The newest heavyweight is Meituan’s LongCat-2.0 (June 29, MIT), a 1.6-trillion-parameter model at 59.5% SWE-Bench Pro trained entirely on Chinese chips. For a clearly commercial license, Kimi K2.7 Code (June 12, Modified MIT) is the newest coding-first leader; DeepSeek V4-Pro (MIT, II 52) tops open models on agentic real-world work; GLM-5.2 (June 13, MIT) runs a 1M-token context for long-horizon autonomous coding; and NVIDIA Nemotron 3 Ultra (June 4, OpenMDW) is the most capable model under a fully permissive license at II 48. DeepSeek V4-Flash is the cheapest 1M-context open model at $0.14 / $0.28 per million tokens.
Qwen 3.7 Max is Alibaba’s flagship API model, launched May 20, 2026 at the Alibaba Cloud Summit in Hangzhou. It scores Intelligence Index 57 on Artificial Analysis (top 10 globally, tied with Claude Opus 4.7, Gemini 3.1 Pro, and GPT-5.5 (medium)), 92.4 on GPQA Diamond, 97.1 on HMMT 2026 February, and 80.4 on SWE-Verified. List API pricing is $2.50 / $7.50 per 1M tokens with a 1M-token context window, plus $0.25 cached input (a 90% cache discount); a 50% launch promotion is still active into July, cutting it to $1.25 / $3.75 (cached input $0.125). Compared to GPT-5.5 at $5 / $30, Qwen 3.7 Max is half the input cost and a quarter of the output cost, but GPT-5.5 still leads on overall Intelligence Index (59-60) and on FrontierMath. For cost-sensitive agentic and long-context work where you want frontier-adjacent quality, Qwen 3.7 Max is the value pick.
The GPT-5.5 family launched April 23, 2026. The headline change is factual reliability: on a selected set of user-flagged conversations, OpenAI reports that GPT-5.5’s individual claims were 23% more likely to be factually correct than GPT-5.4’s, with full responses containing a factual error about 3% less often, plus faster response times across all tiers and a refreshed memory system. The consumer default is GPT-5.5 Instant (free with limits), and the gpt-5.5 API model runs $5 / $30 per 1M tokens. GPT-5.5 Pro is the higher-reasoning variant, available inside ChatGPT Pro at $100/month, and leads FrontierMath Tier 4 at 39.6%. Its successor GPT-5.6 reached general availability on July 9, 2026.
Not on every benchmark, but it is still the best default. GPT-5.6 became ChatGPT’s default on July 9 and leads everyday chat, with GPT-5.5 the proven fallback for factual reliability and reports. Claude Opus 4.8 is the better pick for coding and long agentic tasks and holds #1 on the Intelligence Index at 61. Gemini 3.1 Pro is the better pick for accuracy and research. Gemini 3.5 Flash is the better pick for price-performance. Qwen 3.7 Max is the better pick for cost-effective frontier work. ChatGPT remains the most polished consumer product overall and the natural starting point if you only pay for one model.
Gemini Spark is Google’s first 24/7 cloud-resident AI agent, launched at Google I/O on May 19, 2026 and exclusive to the Google AI Ultra plan, which Google restructured at I/O to include a $100/month entry tier and a $200/month top tier. Spark is built on Gemini base models with Google’s Antigravity harness on a Google Cloud VM, integrates with Gmail, Google Docs, and other Google Workspace apps, and can interact with Chrome and Android’s Halo system on the device side. It is worth the spend for users who have repeatable long-running workflows (inbox triage, research roll-ups, scheduled tasks). For one-off tasks, Claude Cowork at $20/month covers most desktop-agent needs.
On API pricing per million tokens, Gemini 3.5 Flash at $1.50 / $9.00 is the cheapest closed frontier-class model (it beats Gemini 3.1 Pro on coding and agent benchmarks at ~40% lower cost). MiniMax M3 at around $0.60 per million input tokens is the cheapest open-weight frontier coder, with LongCat-2.0 as a strong MIT alternative. Qwen 3.7 Max at $1.25 / $3.75 on its current promo ($2.50 / $7.50 list) is the cheapest at the top Intelligence Index tier, and for an open-weight model with a 1M context, DeepSeek V4-Flash at $0.14 / $0.28 is the cheapest by an order of magnitude. Note that OpenAI applies higher long-context rates to GPT-5.5 for prompts over 272K input tokens ($10 / $45 for GPT-5.5, $60 / $270 for GPT-5.5 Pro).
ChatGPT Free now defaults to GPT-5.6 (with GPT-5.5 still available) under usage limits. Gemini Free runs Gemini 3.5 Flash in the Gemini app and Google AI Studio. Claude Free runs Claude Sonnet 5 (the new default) with daily limits. DeepSeek Chat runs DeepSeek V4 free on the DeepSeek website. Grok has a limited free consumer plan (X Premium is a paid add-on). Qwen 3.5, NVIDIA Nemotron 3 Ultra, MiniMax M3, LongCat-2.0, Kimi K2.6, DeepSeek V4, and GLM-5.2 are open-weight and free to self-host. Qwen 3.7 Max is not free: it is API-only with no consumer chat front-end.
Claude Fable 5, back online since July 1, is the best for coding at 80.3% on SWE-Bench Pro, the highest of any usable model, and is built for long-horizon agentic runs at $10 / $50. Claude Opus 4.8 is the everyday-value pick right behind it at $5 / $25, the favourite inside Cursor and Claude Code. GPT-5.5 is the proprietary alternative at 58.6% SWE-Bench Pro and 82.7% Terminal-Bench 2.0. Gemini 3.5 Flash is the price-performance pick at 76.2% Terminal-Bench 2.1 and $1.50 / $9.00 per 1M tokens. On open weights, LongCat-2.0 (June 29, MIT, 59.5% SWE-Bench Pro) and MiniMax M3 (59% SWE-Bench Pro, ~$0.60 input) are the frontier picks, and Microsoft’s MAI-Code-1-Flash is the budget IDE pick. DeepSeek V4, Kimi K2.6, and GLM-5.2 round out the open-weight coders.
Claude Sonnet 5, launched June 30, 2026, is the best for writing style and instruction-following: it scores 1,618 on GDPval-AA v2 to edge Opus 4.8 (1,615, both behind Fable 5’s 1,783) and is the free and Pro default on claude.ai at $2 / $10 introductory pricing (then $3 / $15). GPT-5.5 is the alternative for fact-anchored business writing (improved factual reliability vs GPT-5.4). Gemini 3.5 Flash is the price-performance pick for bulk content at $1.50 / $9.00.
Gemini 3.1 Pro is the best for accuracy and research at 94.3% GPQA Diamond, 44.4% Humanity’s Last Exam, and 77.1% ARC-AGI-2, with native Google Search grounding for live factual answers. Qwen 3.7 Max is the value runner-up at 92.4 GPQA Diamond. GPT-5.5 Pro is the hallucination-sensitive runner-up. Gemini 3.5 Pro, cleared for a July launch, could take the top spot when it ships.
ChatGPT Images 2.0 is the best for images with readable text, multilingual scripts, and infographic-style output. Google Nano Banana Pro (Gemini 3 Pro Image) is the best for photoreal portraits and products. Reve 2.0 is #2 on the Arena text-to-image leaderboard with native 4K and layout-based editing. ByteDance’s new Seedream 5.0 Pro (July 8) is a strong multilingual-text and region-precise-editing option, though it has no independent benchmarks yet and is accessed via BytePlus and Magnific rather than a consumer app. Midjourney v8 is the best for stylized art, and Grok Imagine is the only frontier model that allows Spicy Mode adult content.
Google Veo 3.1 is the best for AI video after OpenAI retired the Sora 2 consumer app on April 26, 2026 (its API runs until September 24, 2026). Kling 3.5 is the runner-up for fast iteration, Runway Gen-4 is the runner-up for cinematic control.
GPT-5.6 Sol is OpenAI’s new flagship for abstract math and science, included in ChatGPT Pro at $100/month; OpenAI has not yet published Sol’s FrontierMath score, so GPT-5.5 Pro’s verified 39.6% on FrontierMath Tier 4 remains the cited OpenAI mark. Qwen 3.7 Max leads competition math at 97.1 HMMT 2026 February and 44.5 Apex, at a fraction of the cost on API. Claude Opus 4.8 is the alternative for long agentic reasoning, using adaptive thinking and an effort parameter.
Grok 4.5, xAI’s public flagship since July 8, is the best for unfiltered, opinionated, on-trend creativity, with permissive guardrails and native X grounding plus Opus-class reasoning, at $30/month SuperGrok (Grok 4.3 is the cheaper fallback). Claude Opus 4.8 is the alternative for structured long-form creative work, Gemini 3.1 Pro is the alternative for multimodal creative.
Fello AI is an AI chatbot for Mac, iPhone, and iPad that lets you use all top AI models like ChatGPT, Claude, Gemini, Grok, and DeepSeek in one app, with models updated regularly so you always have the latest. It is $9.99/month with a 4.7-star rating across 25,000+ reviews.
We update this page at least monthly and within 24-48 hours of any major model launch.
Připojte se k tisícům fanoušků a profesionálů v oboru umělé inteligence, kteří mohou využít exkluzivní tipy a poznatky předních odborníků v oboru.
