The Ultimate Claude Model Comparison: Claude 4 to Opus 5.5

Anthropic shipped 20 Claude models between May 2025 and October 2026. The newest flagship, Claude Opus 5.5, costs $4 / $20 per million tokens, less than a third of Claude Opus 4's launch price. This Claude models comparison maps every version from Claude 4 to the 5.5 family. The ranking has a twist: on the Artificial Analysis Intelligence Index v4.3.2, the $2 / $10 Sonnet 5.5 scores 56, above the $10 / $50 Fable 5.1 at 53.

The lineup now has four classes. Haiku is the cheap, fast tier. Sonnet is the everyday tier, and Opus is the flagship. The Mythos class sits above Opus as two names for one model: Fable for everyone and Mythos for vetted security and science teams. This guide covers release dates, API prices, benchmark progression, which models each Claude plan includes, and a plain recommendation for every job. For the short news version of what shipped in 2026, read our Claude 5 explainer.

The Key Takeaways

• Claude Opus 5.5 (September 22, 2026) is the model Anthropic's docs tell most workloads to start with. It costs $4 / $20 per million tokens and scores 58 on the Artificial Analysis Intelligence Index v4.3.2, the highest of any Claude model.
• Claude Sonnet 5.5 (September 28, 2026) is the value pick: $2 / $10, a score of 56, and 70.6% on Terminal-Bench 4.0, against 10.3% for Sonnet 5.
• Claude Haiku 5.5 (October 7, 2026) starts at $0.10 / $0.50 for prompts up to 100,000 tokens, which Anthropic says makes it about 75% cheaper to run than Haiku 4.5.
• Claude Fable 5.1 (September 1, 2026) is the most capable model open to all customers at $10 / $50. Claude Mythos 5.1 is the same model with fewer safeguards, open only to vetted US organizations.
• Every current model has a 1M-token context window and 128K-token output. The Claude Free plan includes Sonnet and Haiku, Opus needs Pro ($20 a month) or above, and Fable runs on usage credits on Pro.

Claude Models Comparison: The Current Lineup at a Glance

Anthropic's current Claude lineup is Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, with Mythos 5.1 available only through verification programs. All four general models accept text and images, share a 1M-token context window and a 128K-token output limit, and carry a June 2026 knowledge cutoff, according to Anthropic's models overview.

Speed labels and default effort come from Anthropic's comparison table. Effort defaults differ inside Anthropic's own apps. Sonnet 5.5 runs at Medium in Claude Code and the Claude apps. Fable 5.1 runs at Medium on claude.ai but High in Claude Code. The index scores are Artificial Analysis readings at maximum effort, taken on October 10, 2026.

How the Four Claude Classes Differ

Claude's four classes trade intelligence for speed and price, and each newer generation pushes the cheaper class closer to the one above it. Haiku handles high-volume, latency-sensitive work. Sonnet balances speed and intelligence for daily tasks. Opus is the flagship for long, hard work. The Mythos class, which Anthropic describes as "a tier of Claude models that sit above our Opus class in capability", arrived in April 2026.

The Mythos class ships under two names. Anthropic says Fable comes "from the Latin fabula", akin to the Greek mythos. It adds that "the safeguards are what distinguish the two models." Claude Fable 5.1 is the version anyone can buy. Claude Mythos is the same weights with cybersecurity or biology safeguards lifted for vetted organizations. When Fable's classifiers flag a cyber, biology or distillation request, the answer comes from an Opus model instead, and the user is told.

The gap between classes is now narrow. Anthropic says Opus 5.5 "performs at the level of Claude Fable 5.1 on most work" at 40% of Fable's token price. Sonnet 5.5 scores just two Elo points below Opus 5.5 on GDPval-AA v2.1. Our Claude Opus vs Sonnet comparison covers that pair head to head, and our Claude Fable vs Opus guide covers the top two classes.

Every Claude Model Since Claude 4: Release Timeline

Claude has gone from two models in May 2025 to 20 by October 2026, and only three of those 20 have been retired. The status column comes from Anthropic's model deprecations page, which commits to at least 60 days' notice before any retirement.

Anthropic names Sonnet 5.5 as the replacement for Sonnet 4.5, so anyone still calling claude-sonnet-4-5 has until November 30, 2026 to switch. Among active models, the earliest retirement floors belong to Haiku 4.5 (October 15, 2026) and Opus 4.5 (November 24, 2026). A floor is a minimum, not a shutdown date: each retirement still needs a deprecation notice first.

Claude 4 and 4.1: The Foundation (May to August 2025)

The Claude 4 generation launched on May 22, 2025 with Opus 4 and Sonnet 4, both hybrid models with a near-instant mode and an extended thinking mode. Opus 4 scored 72.5% on SWE-bench Verified and Sonnet 4 scored 72.7%. Anthropic called Opus 4 "the best coding model in the world." Pricing was $15 / $75 for Opus 4 and $3 / $15 for Sonnet 4, and Sonnet 4 was also open to Free users. The same launch made Claude Code generally available. Our Claude 4 launch coverage has the original details.

Opus 4.1 followed on August 5, 2025 at the same price, lifting SWE-bench Verified to 74.5%. All three models are now retired, so none of them appears in the Claude apps or answers on the API.

The Claude 4.5 Generation: Sonnet, Haiku and a Cheaper Opus

The Claude 4.5 generation made Opus affordable and gave Haiku near-frontier coding. Three models shipped in eight weeks, between September 29 and November 24, 2025.

Claude Sonnet 4.5

Claude Sonnet 4.5 launched on September 29, 2025 at $3 / $15 and scored 77.2% on SWE-bench Verified. It led OSWorld computer use at 61.4%, up from Sonnet 4's 42.2% four months earlier. Anthropic reported it holding focus "for more than 30 hours on complex, multi-step tasks." Sonnet 4.5 is now deprecated and retires on November 30, 2026.

Claude Haiku 4.5

Claude Haiku 4.5 arrived on October 15, 2025 at $1 / $5, with 73.3% on SWE-bench Verified. Anthropic pitched it as Sonnet 4-level coding "at one-third the cost and more than twice the speed." It runs a 200K-token context window and remains on the API as a legacy model.

Claude Opus 4.5

Claude Opus 4.5 shipped on November 24, 2025 and cut the Opus price from $15 / $75 to $5 / $25, a two-thirds reduction. It also introduced the effort parameter. At medium effort, Opus 4.5 matched Sonnet 4.5's best SWE-bench Verified score while using 76% fewer output tokens. Our Opus 4.5 launch article covers the release.

Claude 4.6 to 4.8: 1M Context, Adaptive Thinking and Mythos Preview

Between February and May 2026, Claude gained a 1M-token context window, adaptive thinking, finer effort controls and its first Mythos-class model. Every Opus in this stretch kept the $5 / $25 price.

Claude Opus 4.6 and Sonnet 4.6

Claude Opus 4.6 (February 5, 2026) was the first Opus with a 1M-token context window, in beta at launch, and it supports outputs of up to 128K tokens. It introduced adaptive thinking, where Claude decides when deeper reasoning helps, plus four effort levels: low, medium, high and max. On the 8-needle 1M variant of MRCR v2, a long-context retrieval test, Opus 4.6 scored 76% where Sonnet 4.5 scored 18.5%.

Claude Sonnet 4.6 (February 17, 2026) brought the 1M window to Sonnet at $3 / $15 and became the default model for Free and Pro users at launch. In Claude Code testing, users preferred it to Opus 4.5 59% of the time.

Claude Mythos Preview and Opus 4.7

Claude Mythos Preview was announced on April 7, 2026 through Project Glasswing, a limited program for cyber defenders and critical software maintainers. It was the first Mythos-class model, and it never went on general sale.

Claude Opus 4.7 (April 16, 2026) was the first model to ship with automatic cyber safeguards, a test bed for a wider Mythos release. It added the xhigh effort level between high and max. It also reads images up to 2,576 pixels on the long edge, more than three times the pixels of earlier Claude models. Opus 4.7 also introduced a new tokenizer. Anthropic says the same input "can map to more tokens", roughly 1.0 to 1.35 times as many, so an unchanged price per token became a slightly higher price per task.

Claude Opus 4.8

Claude Opus 4.8 (May 28, 2026) put an effort control next to the model picker on claude.ai and Cowork for every plan. Anthropic says it is "around four times less likely" than Opus 4.7 to let flaws in its own code pass unremarked. Fast mode, about 2.5 times the speed, launched at $10 / $50. Opus 4.8 still matters today: when safeguards on Fable 5.1, Opus 5.5 or Opus 5 flag a cybersecurity request, Opus 4.8 is the model that answers.

The Claude 5 Generation: Fable, Mythos, Sonnet 5 and Opus 5

The Claude 5 generation brought Mythos-class intelligence to general users and cut the Sonnet price to $2 / $10. It also produced the most turbulent month in Claude's history.

Claude Fable 5 and Mythos 5

Claude Fable 5 and Claude Mythos 5 launched on June 9, 2026 at $10 / $50, less than half the price of Mythos Preview. Anthropic called Fable 5 "a Mythos-class model that we've made safe for general use," and said over 95% of Fable sessions involved no fallback to another model. Anthropic launched both with mandatory 30-day data retention for Mythos-class traffic, a rule Fable 5.1 later relaxed for eligible enterprise customers.

Three days later, on June 12, the US government issued an export control directive covering both models, and Anthropic suspended access for all users. The controls were lifted on June 30, and Fable 5 returned on July 1. A promotion let subscribers spend up to 50% of their weekly limits on Fable 5 until July 19. After that, Fable became a standard part of Max plans and Team premium seats, while Pro moved to pay-as-you-go usage credits.

Claude Sonnet 5

Claude Sonnet 5 (June 30, 2026) is "close to that of Opus 4.8, but at lower prices," per Anthropic. It launched at an introductory $2 / $10, with $3 / $15 scheduled for September 1. On August 10, Anthropic cancelled that rise and made $2 / $10 permanent.

Claude Opus 5

Claude Opus 5 (July 24, 2026) "comes close to the frontier intelligence of Claude Fable 5 at half the price," Anthropic said, at $5 / $25. Anthropic made it the default model on Max and called it the strongest model on Pro at launch. On ARC-AGI 3, Anthropic reported a score three times the next-best model's. Opus 5 scores 51 on the index, one point above Fable 5.

Claude Fable 5.1 and Mythos 5.1

Claude Fable 5.1 and Mythos 5.1 (September 1, 2026) keep the $10 / $50 price but cut cache reads by 75% to $0.25 per million tokens. Anthropic estimates Fable 5.1 costs about 25% less than Fable 5 on typical workloads and up to about 45% less on heavily agentic ones. Its cyber safeguards intervene about 60% less often per Claude Code session. Mythos 5.1 is open only to a set of US organizations through the Cyber Verification Program and the Life Sciences Verification Program.

The Claude 5.5 Family: Opus 5.5, Sonnet 5.5 and Haiku 5.5

The Claude 5.5 family shipped all three classes in 15 days, from September 22 to October 7, 2026, and each one costs less to run than the model it replaces.

Claude Opus 5.5

Claude Opus 5.5 costs $4 / $20, 20% less per token than Opus 5, with cache reads at $0.20. Anthropic says it runs 40% cheaper than Opus 5 on typical workloads. Output arrives more than 30% faster. It scores 66.4% on Terminal-Bench 4.0 and 1846 Elo on GDPval-AA v2.1, ahead of Fable 5.1 on both. Opus 5.5 is the first Opus to launch with Fable-class safeguards. It also can no longer run with thinking switched off. Our full Opus 5.5 review covers who gets it and on which plan.

Claude Sonnet 5.5

Claude Sonnet 5.5 keeps the $2 / $10 price, runs more than 30% faster than Sonnet 5, and costs up to 30% less per task. On October 7, Anthropic halved its cache-read price to $0.10 per million tokens, cutting most agentic work by about another 20%. It is the first Sonnet with cyber safeguards; flagged requests fall back to Sonnet 5. The Sonnet 5.5 launch guide explains how the effort setting changes the bill.

Claude Haiku 5.5

Claude Haiku 5.5 is Anthropic's fastest model at standard speed and its first Haiku with an adjustable effort setting. Its price depends on prompt length: $0.10 / $0.50 up to 100,000 tokens and $0.50 / $2.50 above. Anthropic says 90% of Haiku 4.5 requests fell under that threshold. Haiku 5.5 scored 39.2% on Terminal-Bench 4.0, where Haiku 4.5 scored 0.0%. Read our Haiku 5.5 guide for when it beats Sonnet.

Each figure comes from that model's Anthropic launch post. The Fable 5.1 numbers come from the comparison table in the Opus 5.5 post; Fable 5.1's own launch post reported a slightly lower 65.0% on Humanity's Last Exam. The index scores come from Artificial Analysis. Opus 5.5's Terminal-Bench result is at xhigh effort. Anthropic notes that Fable and Opus 5.5 were scored with production safeguards on, which "likely reduces" their results. Artificial Analysis's own Terminal-Bench 4.0 runs came out lower, at 64% for Sonnet 5.5 and about 60% for Opus 5.5, but kept Sonnet 5.5 ahead.

Claude Models Comparison by Benchmark: How Much Smarter Each Version Got

On the Artificial Analysis Intelligence Index v4.3.2, Claude climbed from 21 for Sonnet 4.5 to 58 for Opus 5.5 in under 12 months. The index combines ten evaluations, including Terminal-Bench 4.0, GDPval-AA v2.1, Humanity's Last Exam and AA-Briefcase, so it puts every generation on one scale.

What the Index Shows

The index shows each new Sonnet overtaking the previous Opus, Haiku 5.5 passing Opus 4.8, and Fable 5.1 trailing both 5.5 flagships. Sonnet 4.6 at 30 beat Opus 4.5 at 29, and Sonnet 5.5 at 56 beats Opus 5 at 51. Haiku 5.5 at 43 edges Opus 4.8 at 42, a model that costs 50 times as much per input token on prompts up to 100,000 tokens.

The Fable result needs context. Artificial Analysis scores Fable "with fallback", so tasks its safeguards flag are answered by another model. Anthropic's own position is that Opus 5.5 matches Fable 5.1 on most work, not that it is clearly smarter.

Artificial Analysis rescales the index often. Version 4.2 and version 4.3 landed three days apart in September 2026, and both moved every score. Compare these numbers only with other v4.3.2 readings.

Claude API Pricing Compared

Claude API prices range from $0.10 / $0.50 per million tokens for Haiku 5.5 to $10 / $50 for Fable 5.1, and the Batch API halves both input and output. Anthropic's pricing page lists these rates for the models still on sale:

Two rules change the real bill. First, Claude 4.6 and later models bill the full 1M-token window at the standard rate, except Haiku 5.5, which charges the higher band once a prompt passes 100,000 tokens. Second, every model from Opus 4.7 onward uses a newer tokenizer. Anthropic's pricing page says it "produces approximately 30% more tokens for the same text." Sonnet 4.6 still uses the previous tokenizer, so one Sonnet 4.6 token covers more text than one token on the newer models. Fast mode on Opus 5.5 costs $8 / $40. Our Claude pricing breakdown covers plan prices and hidden fees in full.

Which Claude Models Your Plan Includes

The Claude Free plan includes Sonnet and Haiku. Opus needs Pro or above, and Fable is capped or credit-only on Pro, Max and Team. Here is the matrix from Anthropic's pricing page on October 10, 2026:

Pro costs $17 a month billed annually ($200 up front) or $20 monthly. On Team plans, premium seats get Fable at up to 50% of weekly limits, and standard seats use usage credits. Usage credits bill at standard API rates once a paid plan's included usage runs out. From the week of October 7, Max 5x subscribers also get $100 a month in Claude Platform API credits, Max 20x get $200, and Team plans get up to $500 pooled. Our Claude Pro vs Max comparison and Claude usage limits guide explain how much each tier lasts.

Which Claude Model Should You Use?

For most work, use Claude Opus 5.5 if your plan includes it and Claude Sonnet 5.5 if it does not. Anthropic's own model guide says "most workloads start with Claude Opus 5.5" and reserves Fable 5.1 for cases where Opus 5.5 at xhigh or max effort "still fall short."

Before switching models, try changing effort. Anthropic's guide calls tuning effort "often a better lever than switching models", and Opus 5.5 at its default Medium setting beat GPT-6 Astra's top FrontierCode score in Anthropic's tests. If Sonnet 5.5 struggles, raise it to High before paying for Opus.

Use Claude Without Picking a Plan Tier

The plan matrix is where most people get stuck. Opus sits behind a $20 Pro plan, Fable sits behind usage credits or a $100 Max plan, and the API means tracking effort levels and cache rates on top. Fello AI is a native Mac, iPhone and iPad app that puts Claude behind one subscription, alongside ChatGPT, Gemini, Grok, DeepSeek, Perplexity, Kimi, GLM and Qwen. It costs $11.99 a month, with a free tier to try first and a 4.7-star rating across 27,000+ reviews. You can also ask a rival model the same question in the same conversation.

Conclusion

The Claude lineup in October 2026 rewards going one class down. Sonnet 5.5 at $2 / $10 scores 56 on the Artificial Analysis index, ahead of every Opus before 5.5 and ahead of Fable 5.1. Opus 5.5 at $4 / $20 is the strongest Claude on that index and the model Anthropic tells most API customers to start with. Fable 5.1 is now a specialist tool. Haiku 5.5 makes high-volume work cheap enough to run at scale. Start on Sonnet 5.5, move to Opus 5.5 when the work gets hard, and pay Fable prices only when Opus at high effort has already failed. For how Claude's rivals line up, see the ChatGPT model comparison and the Gemini model comparison.

FAQ