Best Open Source AI Models in 2026, Ranked and Compared

The best open source AI models changed hands again in November. Moonshot published the weights for Kimi K3.1 Pro on November 2, 2026, and Artificial Analysis now scores it 46.5 on its Intelligence Index v5.0.1, the highest of any open-weights model, under Moonshot’s own licence rather than MIT. It is also the top open model on Arena’s WebDev board. The top closed model, Claude Fable 5.2, sits at 64.6. Under a plain MIT license, DeepSeek’s V4.1 Flash leads at 32.6, GLM 5.3 Flash is the best MIT-licensed pick for most people, Xiaomi’s MiMo-V2.6-Pro is the cheapest per finished task, and Gemma 4 still runs on a laptop. Our full DeepSeek vs Claude comparison puts one open lab against a closed one side by side.

This guide ranks the open source AI models worth your time in 2026, what each one is genuinely best at, and how to actually run them without a rack of GPUs. We ranked the picks on independent benchmark data, license terms, and real accessibility, because most “best open source AI” lists are written by companies selling you the hardware to run them. We sell an app, not GPUs, so this ranking has no horse in that race. For a closer look at how an open model holds up against the paid flagships, see GLM head-to-head with the closed-source flagships.

If the “open source” label itself feels fuzzy, our explainer on what open source AI actually means breaks down open weights, licensing, and why “open” does not always mean what you think.

The Key Takeaways

• Kimi K3.1 Pro is the highest-scoring open model on the Artificial Analysis Intelligence Index v5.0.1 at 46.5, open since November 2 under Moonshot’s own licence and the top open model on Arena’s WebDev and Agent boards, at 2.78 trillion parameters. Under a plain MIT license, DeepSeek V4.1 Flash leads at 32.6, and MiMo-V2.6-Pro (29.9) is the cheapest per finished task at about $0.26.
• GLM 5.3 Flash is the right pick for most people: 27.8 on the index, within five points of the MIT leader, a 1M-token context, native image and video input, and a clean MIT license on a 321B model you can actually rent hardware for. The older GLM 5.2 is deprecated and has no score on the current index.
• DeepSeek V4.1 Flash (MIT) is the second-highest-scoring open model, after Kimi K3.1 Pro and ahead of the full GLM 5.3 at 31.7, at about $0.60 per index task.
• Most “open source” models are actually open-weight, the weights are free but the training data stays private. True open-source AI is rarer than the label suggests, and several of the strongest models here ship under custom licenses rather than MIT or Apache 2.0.
• You do not need a GPU farm for most of this list. Gemma 4 12B runs on a modern laptop, Qwen 3.8 27B fits a well-equipped Mac, and apps route the biggest models to your Mac or iPhone with zero local hardware.

What Counts as an Open Source AI Model

An open source AI model publishes its weights, and ideally its code and training data, so anyone can download, run, modify, and redistribute it. That is the promise. The reality is more nuanced, and it matters before you pick one.

Most models everyone calls “open source” are really open-weight. You get the trained weights for free, but the training data and full recipe stay locked away. The Open Source Initiative’s official definition requires far more transparency than that, which is why almost no frontier model qualifies as fully open. We flag the distinction throughout, because a “Modified MIT” or custom community license can carry commercial strings that a true MIT or Apache 2.0 license does not.

The practical takeaway is simple. Open weights mean you can run the model privately, fine-tune it, and avoid per-token API fees. They do not always mean unrestricted commercial use, so the license column in our table below is just as important as the benchmark scores.

Best Open Source AI Models 2026 at a Glance

Here is the full ranking. We picked these ten on independent benchmark performance, recency, license freedom, and how realistic they are to actually use. The index column is the Artificial Analysis Intelligence Index v5.0.1, read on November 10, 2026. Artificial Analysis rebuilt the index as version 5 on October 27, so scores from earlier versions are not comparable with these: Kimi K3.1 Pro scored 55.1 on the old v4.3.2 and scores 46.5 now, the same model on a different ruler. Models marked “not scored” have no version 5 result yet. Chinese labs hold the top five places. The strongest US-built open models, Meta’s Muse Spark Open and Thinking Machines’ Inkling, get their own section below.

The Best Open Source AI Models, Ranked

Here is each pick in detail, in order, with what it is genuinely best at, the benchmarks that earned its spot, and the license you actually get. We start with the strongest all-rounders and work down to the most efficient and the most local.

1. Kimi K3.1 Pro, the Highest-Scoring Open Model

Kimi K3.1 Pro from Moonshot AI became a download on November 2, 2026, ten days after its launch, and it ties its predecessor Kimi K3 as the biggest open model you can get. It keeps K3’s 2.78 trillion parameters in a Mixture-of-Experts design that activates 104 billion per token, with a 1-million-token context window. The nearest rival on size is Alibaba’s Qwen 3.8 flagship at 2.45 trillion.

The independent boards agree on it. Artificial Analysis scores K3.1 Pro 46.5 on Intelligence Index v5.0.1, the highest of any open-weights model and 13.9 points clear of the next one, DeepSeek V4.1 Flash. It also tops Artificial Analysis’s Terminal-Bench 4.0 coding-agent run at 67.2%, ahead of every closed model. On Arena it is the highest open entry on the WebDev board, 10th overall with only closed models above it, and on the Agent board. Kimi K3, which held this spot until November, has no score on the current index, because Artificial Analysis lists it as superseded by K3.1 Pro.

Two things to know before you plan around it. The weights ship under Moonshot’s own Kimi K3.1-Pro License, which lets you use, modify and sell the model but asks hosted-service businesses with more than $20 million in revenue to sign a separate agreement, and very large products to show the model’s name. The other catch is hardware: at 2.78 trillion parameters this is a server model and not a laptop one, and Artificial Analysis puts one index task at $3.68, because the model writes long answers. Our Kimi K3 breakdown covers the predecessor and its licence.

2. DeepSeek V4.1 Flash, the Highest-Scoring MIT-Licensed Model

DeepSeek V4.1 Flash replaced V4-Flash on September 10, 2026, and it is now the strongest open model DeepSeek makes and the highest-scoring one under a plain MIT license. The model card describes a multimodal Mixture-of-Experts model with 552 billion backbone parameters that activates only 8 billion per token while reading input and 16 billion while writing, with native image input, a 1-million-token context and an MIT license. Meituan’s open-sourced LongCat-2.0 is another 1M-token option, built specifically for agentic coding.

Its 17 October build, V4.1-Flash-1017, scores 32.6 on Intelligence Index v5.0.1 at about $0.60 per index task, ahead of the full GLM 5.3 at 31.7 and MiMo-V2.6-Pro at 29.9. The bigger V4-Pro (1.6 trillion total, 49 billion active) is now the weaker of the two at 23.1 and has no image input, so its April launch scores, 80.6 on SWE-Bench Verified and 93.5 on LiveCodeBench, describe a model DeepSeek has already overtaken itself. Read the full DeepSeek V4 breakdown for both models and the current peak and off-peak API prices.

3. GLM 5.3 Flash, the Best Open Source Model for Most People

GLM 5.3 Flash is the open model most teams should actually run. Released by Z.ai on August 26, 2026, it is a 321-billion-parameter Mixture-of-Experts model that activates about 18 billion per token, reads a 1-million-token context, takes images and video as well as text, and ships under a clean MIT license. It scores 27.8 on Intelligence Index v5.0.1 at about $0.58 per task, within five points of DeepSeek V4.1 Flash on well under two-thirds of its parameters.

It replaces GLM 5.2, which held this slot until September. GLM 5.2 is now deprecated and has no score on the current index, and Z.ai charges the same $1.40 / $4.40 per million tokens for it as for the full GLM 5.3, which scores 31.7. GLM 5.3 is the stronger model and its weights are on Hugging Face, but they ship under a custom license rather than MIT, which is why the Flash model gets the recommendation. The rest of Zhipu’s GLM family is covered in our explainer, and GLM 5.5 is rumoured for later in 2026.

4. MiMo-V2.6-Pro, Best Value per Finished Task

MiMo-V2.6-Pro from Xiaomi took the top of the open field on September 21, 2026, the day its weights went up on Hugging Face. Xiaomi’s model card describes a sparse Mixture-of-Experts design with 1.02 trillion total and 42 billion active parameters, a 1-million-token context, and native text, image, video and audio input in one model. The license is plain MIT, with no revenue threshold and no attribution clause, which neither of the two open models closest to it on the index can say.

The case for it now is cost. Artificial Analysis scores it 29.9 on Intelligence Index v5.0.1, behind DeepSeek V4.1 Flash and the full GLM 5.3, but one index task costs about $0.26, the lowest of the five highest-scoring open models, against $0.60 for DeepSeek and $3.68 for Kimi K3.1 Pro. The caveat is human preference: nobody had voted on it at launch, and Arena now ranks it only 25th on WebDev, well behind the Kimi models. Our Xiaomi MiMo explainer covers the whole family, its API prices and the smaller V2.6 Flash.

5. Qwen 3.8 27B, the Best Open Model You Can Run on a Mac

Alibaba’s Qwen 3.8 27B is the strongest model on this list that fits on a single well-equipped machine. It is a 27.8-billion-parameter dense model under a clean Apache 2.0 license, released on Hugging Face in August 2026, with native image and video understanding and a 262K-token context that Alibaba says extends to 1 million. It scores 20.1 on Intelligence Index v5.0.1 at its highest setting, more than three times Muse Glimmer’s score. The larger Qwen 3.8 flagship, 2.45 trillion parameters, also has public weights, but it ships under Alibaba’s own license, as does the 180B Qwen3.8-Flash-Next, and neither has a score on the current index yet.

The earlier Qwen 3.6 models are still downloadable, and the 35B-A3B is still the lighter option for modest hardware, but the 27B has moved on. Alibaba’s strongest model overall, Qwen3.8 Max, scores 33.7 and is API-only and closed-weight, so it sits outside this open-source ranking. For one open model that does a bit of everything on hardware you own, Qwen 3.8 27B is the safest bet. If you run it yourself, the Prague lab BottleCap AI publishes ThinkingCap fine-tunes that cut its reasoning tokens by 37% to 46% for under a point of accuracy.

6. Kimi K2.7 Code, Best for Agentic Coding on Less Hardware

Kimi K3.1 Pro and K3 are too big for most teams, so Kimi K2.7 Code is the Kimi most teams can realistically deploy, because it is under two-fifths of their size. It is a 1-trillion-parameter model activating 32 billion per token, with a 256K-token context window under a Modified MIT license, released June 12, 2026 and purpose-built for agents that write, run, and debug across many steps. Moonshot’s hosted K2 API closed to new users on 31 October 2026 and retires by the end of December. You can see what Moonshot’s hosted models cost in our Kimi pricing guide.

One honest caveat. Most of what Moonshot published, including a 62.0 on its own Kimi Code Bench v2 and 81.1 on MCP Mark Verified, comes from in-house benchmarks. The first independent reading was less flattering: Artificial Analysis scored it 25.8 on the older v4.3.2 index, well below the models above it, though that index is not a coding-only test, and it has no score on the current version yet. Treat it as a specialist rather than an all-rounder. Our Kimi K2.7 Code review has the details and the missing-data flags.

7. Muse Glimmer, Meta’s Open Model for Local Agents

Meta’s first open-weight release this year was Muse Glimmer, published on August 10, 2026 under a fully permissive Apache 2.0 license. It is a 30B dense model (about 29.8 billion parameters, including its vision encoder) distilled from Meta’s closed Muse Spark, with a 131K context window and image input, and Meta built it for agents that use tools and recover from their own mistakes rather than for chat. There is no Llama 5. Meta’s last Llama generation was Llama 4 in April 2025, and its 2026 models ship under the Muse name, with both Muse Spark and the newer flagship Muse Wonder kept closed. Meta kept its promise of open weights for Muse Spark on 15 October 2026 with Muse Spark Open, an 816B-parameter model (42B active) under Meta’s own Muse Model License that Meta says needs two 512 GB Mac Studios to run locally. It has its own entry with the other US-built open models below.

What makes Glimmer notable is that Meta benchmarked it on Macs itself. Its 4-bit build shrinks the language model to under 20 GB, the smallest official quant is aimed at 24 GB machines, and on Meta’s own numbers it runs at 37.8 tokens per second on an M4 Max and 50.2 on an M5 Max with its DFlash speculative decoding switched on. Ollama ships an Apple Silicon build you can pull with ollama run muse-glimmer:30b-mlx. The independent picture is modest, 6.3 on Intelligence Index v5.0.1, and Meta’s own table is mixed: it beats Gemma 4 31B clearly on MCP Atlas (75.5 against 54.2) and SWE-Bench Pro (51.2 against 36.9) but loses to Qwen 3.6 27B on OSWorld-Verified and Terminal-Bench 2.1. It is a strong local agent model, not a straight upgrade over everything in its size class.

8. Gemma 4, Best Lightweight Model for Local Use

If you want to run AI on your own machine with the least fuss, Gemma 4 from Google DeepMind is the answer. The family spans tiny edge models (E2B, E4B) and a 12B model up to a 31B dense flagship and a 26B MoE variant, all under Apache 2.0, released April 2, 2026.

Despite the small footprint it punches hard on the boards people vote on. Gemma 4 31B sits at 1,451 on Arena’s text board, above MiniMax M3, Inkling and Qwen 3.8 27B, even though Artificial Analysis scores it only 4.4 on its tougher index. Google’s own figures are 85.2% on MMLU-Pro, 89.2% on AIME 2026, and 80.0% on LiveCodeBench v6, while the 12B tier runs comfortably on a modern laptop. It is the best entry point if you are new to local AI, and Google’s official Gemma 4 model card lists exact specs per size.

9. Nex-N2-Pro, Best Fully Permissive Agentic Model

Nex-N2-Pro from Nex AGI earns its spot on license freedom plus genuine capability. It is a 397-billion-parameter MoE model (17B active) built on Qwen 3.5’s largest open model, with a 262K context, shipped under the fully permissive Apache 2.0 license and tuned specifically for agentic work like tool-calling and long multi-step tasks.

Nex AGI reports 80.8 on SWE-Bench Verified and 75.3 on Terminal-Bench 2.1. Artificial Analysis estimated it at 28.2 on the older v4.3.2 index and has not scored it on the current version, and it carries zero commercial restrictions. See our Nex-N2-Pro analysis for how it stacks up against the top closed models.

10. MiniMax M3.1 Flash, Best for Efficiency

MiniMax M3.1 Flash rounds out the list for anyone watching hardware cost. MiniMax launched it on October 19, 2026 and published the weights on Hugging Face later that month. It runs about 428 billion total parameters with only 23 billion active, uses MiniMax Sparse Attention throughout, reads text, images and video, and carries a full 1-million-token context. MiniMax reports 73.7% on DeepSWE v1.1, Artificial Analysis scores it 29.1 on Intelligence Index v5.0.1 at about $0.49 per task, and MiniMax’s API charges a flat $0.15 / $0.60 per million tokens.

It replaces MiniMax M3 in this slot, which now scores 18.9, and keeps the license M3 introduced, the MiniMax Community License: free to self-host, but larger commercial users need a separate agreement, a step back from the Modified MIT terms of earlier versions. It is still a smart pick when you want fast, frontier-adjacent coding on a leaner hardware budget. Our MiniMax model review covers the lineage and the benchmark caveats.

Muse Spark Open and Inkling, the Strongest US Open-Weights Models

Muse Spark Open is the highest-scoring open-weights model built in the US. Meta released it on October 15, 2026: a mixture-of-experts model with 816 billion parameters, 42 billion active per token, a 1M-token context window, and text, image, video and audio input. Artificial Analysis scores it 24.9 on Intelligence Index v5.0.1 at its xhigh setting, at about $2.02 per index task. That is well ahead of every other US open model on this page, but behind the Chinese leaders in the ranking above.

Two catches keep it out of the numbered list. The weights ship under Meta’s own Muse Model License with limited commercial terms, not Apache 2.0 or MIT. The full model is also about 1.5 TB at BF16, so Meta’s own local setup is two M5 Ultra Mac Studios with 512 GB each. It is the closest Meta has come to the open flagship people kept calling Llama 5.

Inkling is the first model from Thinking Machines, the lab founded by former OpenAI CTO Mira Murati, and it landed on July 15, 2026. It is a Mixture-of-Experts transformer with 975 billion total parameters and 41 billion active per token, a context window of up to 1 million tokens, and native reasoning over text, images and audio. The weights ship under a plain Apache 2.0 license, which puts it in the most permissive tier on this page alongside Qwen 3.8 27B, Gemma 4, Muse Glimmer and Nex-N2-Pro.

The headline is national rather than global. On Artificial Analysis Intelligence Index v5.0.1, Inkling scores 12.5, ahead of NVIDIA’s Nemotron 3 Ultra at 11.8, Muse Glimmer at 6.3 and gpt-oss-120b at 3.8, which put the family at the top of the US open field until Muse Spark Open arrived. It is also more than 30 points behind Kimi K3.1 Pro, so even at its peak Inkling led the American field, not the world. On its own scorecard Inkling posts 77.6% on SWE-bench Verified and 73.5% on MMMU Pro.

Where it earns attention is cost. Artificial Analysis puts one Inkling index task at about $1.35, cheaper than Kimi K3.1 Pro or GLM-5.3, though MiMo-V2.6-Pro and GLM 5.3 Flash now score far higher for less.

The more interesting release is the small one. Inkling-Small shipped its full weights on July 30, 2026 at 276 billion total parameters and 12 billion active, under a third of its parent, and Artificial Analysis scores it 12.9, just above the flagship, at about $0.20 per index task. It also edges the larger model on Humanity’s Last Exam (33% against 32%) and GPQA Diamond (89% against 87%). If you want a US-built model on hardware you can realistically rent, Inkling-Small is the more practical half of this pair.

We have left Inkling out of the numbered ranking above because on capability alone it would sit near the bottom of it, while on license freedom it matches the best on this page. Treat it as the default pick when you specifically want a US-built model on a fully permissive license. Tencent’s 770B Hy4 Preview arrived under Apache 2.0 on August 28, 2026 and is listed here rather than ranked, since no independent board has rated it. The official Hy4 that followed on November 4 scores 34.6 on Artificial Analysis, but Tencent has not published its weights.

How We Picked These Open Source AI Models

We ranked on four things, in order. Independent benchmark performance, led by the Artificial Analysis Intelligence Index and Arena’s human-vote boards, with lab-reported tests (SWE-Bench Verified, GPQA Diamond, MMLU-Pro, LiveCodeBench) as supporting evidence. Recency, every model here shipped or updated in 2026. License freedom, with a clear preference for MIT and Apache 2.0 over custom community licenses. And accessibility, because a model you cannot realistically run is not useful to you. Our guide to AI benchmarks explains what each of those tests actually measures.

We also weighted independent verification. Where scores come only from the lab that built the model, as with parts of Moonshot’s Kimi tables and most of Nex AGI’s, we say so, and where Artificial Analysis has not scored a model on the current index we mark it. Self-reported benchmarks are a real problem in open AI right now, and you should discount them until third parties confirm the numbers.

You Do Not Need a GPU Farm to Use These

The biggest myth about open source AI is that running it requires server-grade hardware. It depends entirely on the model. The trillion-parameter giants like Kimi K3.1 Pro and MiMo-V2.6-Pro do need multi-GPU setups for full-precision inference, but plenty of strong open models run on consumer gear.

Gemma 4 12B, the smaller Qwen tiers and Meta’s Muse Glimmer run on a modern laptop or Mac using tools like Ollama or LM Studio, and you can grab the weights straight from Hugging Face. And if you want the biggest models without owning any hardware at all, an app like Fello AI routes leading models straight to your Mac or iPhone, no local install required. Running a model yourself also sidesteps provider outages entirely, a real concern given Claude’s repeated 2026 downtime. Three other systems are worth knowing about even though they are not ranked here. ByteDance’s Seed 2.1 Pro is an unreleased preview that reached #8 on Code Arena: Frontend in June. Sakana AI’s Fugu orchestrates a pool of other models instead of competing as a single set of weights, and its Fugu-Ultra v1.1 update widened that lead on coding and terminal benchmarks. Quantized versions also shrink memory needs dramatically, often at minimal quality cost.

If coding is your main use case, our roundup of the best free AI for coding compares these open models head-to-head on developer tasks specifically.

Conclusion

Kimi K3.1 Pro is the highest-scoring open source AI model, open since November 2 and the top open entry on Arena’s WebDev board, but at 2.78 trillion parameters it is a data-centre model. Under a plain MIT license, DeepSeek V4.1 Flash scores highest. For most people GLM 5.3 Flash is the better choice, with the strongest blend of reasoning, license freedom, and a 1M-token context on hardware you can actually rent. If you want the lowest cost per finished task, go MiMo-V2.6-Pro. If you want to run AI on your own Mac, start with Qwen 3.8 27B or, on lighter hardware, Gemma 4. If you want the strongest US-built open model, Muse Spark Open leads, under Meta’s own license. On a clean Apache 2.0 license, Inkling-Small is the best value in that corner right now. And if you would rather skip the setup entirely, the same frontier open models are available through the best AI models on apps like Fello AI.

The gap between the best open model and the best closed one is 18.1 points on Artificial Analysis Intelligence Index v5.0.1, 46.5 for Kimi K3.1 Pro against 64.6 for Claude Fable 5.2. The best MIT-licensed model, DeepSeek V4.1 Flash, is 32 points back at 32.6, but one index task costs about $0.60 on it against $26.54 on Fable 5.2 at its maximum setting, so “free” is often still the smarter choice. Pick by your use case, check the license, and you will not miss the subscription. Check the release terms too, because GLM 5.3 and its staged weights show that an open line can ship its best model under stricter terms than the one before it.

FAQ