AI Rankings
Best Open-Source LLMs: Open-Weights Models Ranked by Intelligence
Data updated
Top 20 at a glance
Bars start at 140, not zero, so differences are easier to see.
Full ranking
| Rank and model | Epoch Capabilities Index | Likely range | Access (Epoch) | Price per 1M tokens (in / out) | Released |
|---|---|---|---|---|---|
1Kimi K3Moonshot AI · Open weights | 157.9 | 155.1 to 160.6 | Open weights (non-commercial) | $3.00 / $15.00 | 2026-07-16 |
2GLM-5.3Z.ai · Open weights | 156.0 | 154.2 to 158.3 | Open weights (restricted use) | $1.40 / $4.40 | 2026-08-14 |
3DeepSeek V4.1 FlashDeepSeek · Open weights | 155.9 | 153.7 to 158.6 | Open weights (unrestricted) | $0.30 / $1.20 | 2026-09-09 |
4DeepSeek V4 Pro 0813DeepSeek · Open weights | 155.6 | 153.7 to 158.0 | Open weights (unrestricted) | $1.32 / $3.96 | 2026-08-13 |
5Muse Spark OpenMeta · Open weights | 155.1 | 152.7 to 158.0 | Open weights (restricted use) | $1.25 / $4.25 | 2026-10-15 |
6DeepSeek V4 Flash 0731DeepSeek · Open weights | 154.8 | 152.1 to 156.9 | Open weights (unrestricted) | $0.19 / $0.38 | 2026-07-31 |
7GLM-5.3-FlashZ.ai · Open weights | 153.1 | 150.8 to 155.5 | Open weights (unrestricted) | $0.15 / $0.50 | 2026-08-20 |
8GLM-5.2Z.ai · Open weights | 151.9 | 150.0 to 154.1 | Open weights (unrestricted) | $1.40 / $4.40 | 2026-06-16 |
9Qwen 3.8 27BAlibaba · Open weights | 151.7 | 149.8 to 153.7 | Open weights (unrestricted) | $0.45 / $3.00 | 2026-08-14 |
10Kimi K2.6Moonshot AI · Open weights | 150.9 | 149.0 to 152.7 | Open weights (unrestricted) | $0.95 / $4.00 | 2026-04-20 |
11Inkling-SmallThinking Machines · Open weights | 150.3 | 147.2 to 152.1 | Open weights (unrestricted) | $0.47 / $1.20 | 2026-07-15 |
12Kimi K2.7 CodeMoonshot AI · Open weights | 150.1 | 148.2 to 151.8 | Open weights (unrestricted) | $0.95 / $4.00 | 2026-06-12 |
13GLM-5.1Z.ai · Open weights | 149.8 | 147.7 to 151.9 | Open weights (unrestricted) | $1.40 / $4.40 | 2026-04-07 |
14DeepSeek-V4-ProDeepSeek · Open weights | 149.1 | 147.6 to 150.9 | Open weights (unrestricted) | $1.32 / $3.96 | 2026-04-24 |
15InklingThinking Machines · Open weights | 148.4 | 146.0 to 150.2 | Open weights (unrestricted) | $1.00 / $4.05 | 2026-07-15 |
16Kimi K2.5Moonshot AI · Open weights | 148.1 | 146.6 to 149.5 | Open weights (unrestricted) | $0.60 / $3.00 | 2026-01-27 |
17MiniMax-M3MiniMax · Open weights | 146.9 | 142.3 to 150.2 | Open weights (restricted use) | $0.30 / $1.20 | 2026-06-01 |
18Qwen3.5 397B-A17BAlibaba · Open weights | 146.8 | 145.2 to 148.9 | Open weights (unrestricted) | $0.60 / $3.60 | 2026-02-13 |
19MiniMax-M2.5MiniMax · Open weights | 146.6 | 142.2 to 147.9 | Open weights (unrestricted) | $0.30 / $1.20 | 2026-02-12 |
20Qwen3.6 27BAlibaba · Open weights | 146.2 | 144.1 to 147.7 | Open weights (unrestricted) | $0.37 / $3.20 | 2026-04-22 |
21DeepSeek-V4-FlashDeepSeek · Open weights | 146.1 | 143.8 to 148.0 | Open weights (unrestricted) | $0.30 / $1.20 | 2026-04-24 |
22DeepSeek-V3.2DeepSeek · Open weights | 146.1 | 144.3 to 147.4 | Open weights (unrestricted) | $0.28 / $0.40 | 2025-12-01 |
23Nemotron 3 UltraNVIDIA · Open weights | 146.0 | 143.7 to 148.2 | Open weights (unrestricted) | Not measured | 2026-06-04 |
24Kimi K2 ThinkingMoonshot AI · Open weights | 145.8 | 143.2 to 147.5 | Open weights (restricted use) | $0.60 / $2.50 | 2025-11-06 |
25GLM-5Z.ai · Open weights | 145.8 | 143.8 to 147.7 | Open weights (unrestricted) | $1.00 / $3.20 | 2026-02-11 |
26MiniMax-M2.7MiniMax · Open weights | 145.8 | 138.4 to 148.2 | Open weights (non-commercial) | $0.30 / $1.19 | 2026-03-18 |
27DeepSeek-V3.2-ExpDeepSeek · Open weights | 144.9 | 142.4 to 146.8 | Open weights (unrestricted) | $0.27 / $0.41 | 2025-09-29 |
28Qwen3-235B-A22B-Thinking (Jul 2025)Alibaba · Open weights | 143.8 | 141.7 to 145.3 | Open weights (unrestricted) | $0.30 / $2.90 | 2025-07-25 |
29Qwen 3.6 35B-A3BAlibaba · Open weights | 143.7 | 141.1 to 145.8 | Open weights (unrestricted) | $0.20 / $1.10 | 2026-04-14 |
30GLM-4.7Z.ai · Open weights | 143.6 | 141.4 to 145.8 | Open weights (unrestricted) | $0.60 / $2.20 | 2025-12-22 |
31Gemma 4 31B ITGoogle · Open weights | 143.0 | 140.3 to 145.8 | Open weights (restricted use) | $0.14 / $0.40 | 2026-04-02 |
32Qwen3.5-35B-A3BAlibaba · Open weights | 142.5 | 140.4 to 145.0 | Open weights (unrestricted) | $0.25 / $1.25 | 2026-02-24 |
33Gemma 4 26B A4BGoogle · Open weights | 141.9 | 138.9 to 143.4 | Open weights (unrestricted) | $0.15 / $0.60 | 2026-04-02 |
34Mistral Medium 3.5Mistral AI · Open weights | 141.4 | 138.5 to 144.3 | Open weights (restricted use) | $1.50 / $7.50 | 2026-04-28 |
35DeepSeek-R1 (May 2025)DeepSeek · Open weights | 141.1 | 139.1 to 142.7 | Open weights (unrestricted) | $0.75 / $2.45 | 2025-05-28 |
36Kimi K2 (Jul 2025)Moonshot AI · Open weights | 140.0 | 136.3 to 141.9 | Open weights (restricted use) | $0.55 / $2.20 | 2025-07-12 |
37DeepSeek-V3.1DeepSeek · Open weights | 139.8 | 135.7 to 143.3 | Open weights (unrestricted) | $0.56 / $1.68 | 2025-08-21 |
37gpt-oss-120bOpenAI · Open weights | 139.8 | 135.6 to 142.1 | Open weights (unrestricted) | $0.15 / $0.60 | 2025-08-05 |
39Qwen3-30B-A3B-Thinking (Jul 2025)Alibaba · Open weights | 139.6 | 135.4 to 141.0 | Open weights (unrestricted) | $0.90 / $0.90 | 2025-07-30 |
40Qwen3.5-9BAlibaba · Open weights | 139.3 | 136.1 to 140.8 | Open weights (restricted use) | $0.17 / $0.15 | 2026-02-24 |
41Qwen3-235B-A22BAlibaba · Open weights | 139.2 | 134.6 to 140.5 | Open weights (unrestricted) | $0.21 / $0.84 | 2025-04-28 |
42DeepSeek-R1DeepSeek · Open weights | 138.8 | 136.0 to 140.2 | Open weights (unrestricted) | $0.55 / $2.19 | 2025-01-20 |
43Qwen3-235B-A22B-Instruct (Jul 2025)Alibaba · Open weights | 138.8 | 135.5 to 140.8 | Open weights (unrestricted) | $0.22 / $0.88 | 2025-07-25 |
44Qwen3-32BAlibaba · Open weights | 138.4 | 134.7 to 140.2 | Open weights (unrestricted) | $0.10 / $0.38 | 2025-04-29 |
45Qwen3-14BAlibaba · Open weights | 138.2 | 133.2 to 140.0 | Open weights (unrestricted) | $0.08 / $0.24 | 2025-04-29 |
46gpt-oss-20bOpenAI · Open weights | 137.7 | 132.2 to 139.6 | Open weights (unrestricted) | $0.06 / $0.24 | 2025-08-05 |
47QwQ-32BAlibaba · Open weights | 137.5 | 132.6 to 141.8 | Open weights (unrestricted) | $0.20 / $0.45 | 2025-03-05 |
48Qwen3-30B-A3B-Instruct (Jul 2025)Alibaba · Open weights | 137.3 | 132.0 to 139.5 | Open weights (unrestricted) | $0.10 / $0.30 | 2025-07-29 |
49Qwen3-30B-A3BAlibaba · Open weights | 136.1 | 130.9 to 138.7 | Open weights (unrestricted) | $0.095 / $0.39 | 2025-04-29 |
50Qwen3-8BAlibaba · Open weights | 136.1 | 129.9 to 138.3 | Open weights (unrestricted) | $0.04 / $0.14 | 2025-04-28 |
51DeepSeek-V3 (Mar 2025)DeepSeek · Open weights | 135.8 | 132.2 to 137.8 | Open weights (restricted use) | $0.45 / $0.90 | 2025-03-24 |
52Magistral Small 1.0Mistral AI · Open weights | 133.1 | 129.0 to 134.3 | Open weights (unrestricted) | Not measured | 2025-06-10 |
53DeepSeek-V3DeepSeek · Open weights | 132.3 | 128.1 to 135.2 | Open weights (restricted use) | $0.27 / $1.10 | 2024-12-26 |
54Llama 4 MaverickMeta · Open weights | 132.2 | 127.7 to 134.3 | Open weights (restricted use) | $0.22 / $0.79 | 2025-04-06 |
55Phi-4Microsoft · Open weights | 130.5 | 124.7 to 132.0 | Open weights (unrestricted) | $0.098 / $0.32 | 2024-12-12 |
56Gemma 3 27BGoogle · Open weights | 130.0 | 124.5 to 132.0 | Open weights (restricted use) | $0.12 / $0.30 | 2025-03-12 |
57Llama 4 ScoutMeta · Open weights | 129.6 | 125.1 to 131.2 | Open weights (restricted use) | $0.10 / $0.30 | 2025-04-05 |
58Qwen2.5-72BAlibaba · Open weights | 129.0 | 123.5 to 130.7 | Open weights (unrestricted) | $0.37 / $0.40 | 2024-09-19 |
59Llama 3.1-405BMeta · Open weights | 128.8 | 124.4 to 130.6 | Open weights (restricted use) | $2.40 / $2.40 | 2024-07-23 |
60Mistral Large 2 (Nov 2024)Mistral AI · Open weights | 128.6 | 124.0 to 130.6 | Open weights (non-commercial) | $2.00 / $6.00 | 2024-11-18 |
61Mistral Large 2 (Jul 2024)Mistral AI · Open weights | 127.6 | 122.0 to 129.7 | Open weights (non-commercial) | Not measured | 2024-07-24 |
62Mistral Small 3.1Mistral AI · Open weights | 127.5 | 122.3 to 129.3 | Open weights (unrestricted) | Not measured | 2025-03-17 |
63Llama 3.3 70BMeta · Open weights | 127.3 | 122.5 to 129.4 | Open weights (restricted use) | $0.68 / $0.68 | 2024-12-06 |
64Llama 3.1-70BMeta · Open weights | 125.9 | 120.6 to 128.2 | Open weights (restricted use) | $0.72 / $0.72 | 2024-07-23 |
65Llama 3.2 90BMeta · Open weights | 125.5 | 120.1 to 127.2 | Open weights (restricted use) | $0.72 / $0.72 | 2024-09-24 |
66Qwen2-72BAlibaba · Open weights | 125.3 | 119.1 to 126.9 | Open weights (unrestricted) | $0.90 / $0.90 | 2024-06-07 |
67DeepSeek-V2 (MoE-236B, May 2024)DeepSeek · Open weights | 124.8 | 118.8 to 127.8 | Open weights (restricted use) | Not measured | 2024-05-07 |
68Gemma 3 12BGoogle · Open weights | 123.5 | 114.2 to 128.8 | Open weights (restricted use) | $0.07 / $0.13 | 2025-03-12 |
69Llama 3-70BMeta · Open weights | 123.0 | 117.3 to 125.8 | Open weights (restricted use) | $0.62 / $1.77 | 2024-04-18 |
70Gemma 2 27BGoogle · Open weights | 122.1 | 115.5 to 124.2 | Open weights (restricted use) | $0.80 / $0.80 | 2024-06-24 |
71Mixtral 8x22BMistral AI · Open weights | 122.0 | 115.3 to 124.4 | Open weights (unrestricted) | $1.20 / $1.20 | 2024-04-17 |
72phi-3-small 7.4BMicrosoft · Open weights | 121.8 | 114.3 to 124.6 | Open weights (unrestricted) | Not measured | 2024-04-23 |
73phi-3-medium 14BMicrosoft · Open weights | 121.2 | 114.3 to 124.4 | Open weights (unrestricted) | Not measured | 2024-04-23 |
74Gemma 2 9BGoogle · Open weights | 119.8 | 111.8 to 122.3 | Open weights (restricted use) | $0.20 / $0.20 | 2024-06-24 |
75Qwen2.5-Coder-32BAlibaba · Open weights | 119.5 | 110.8 to 124.8 | Open weights (unrestricted) | $0.66 / $0.80 | 2024-09-18 |
76Command R+Cohere · Open weights | 119.2 | 110.3 to 125.0 | Open weights (non-commercial) | $2.75 / $12.50 | 2024-08-30 |
77Mistral NeMoMistral AI · Open weights | 118.7 | 111.1 to 121.4 | Open weights (unrestricted) | $0.13 / $0.15 | 2024-07-18 |
78Qwen2.5-7BAlibaba · Open weights | 118.5 | 109.6 to 121.7 | Open weights (unrestricted) | $0.07 / $0.10 | 2024-09-19 |
79Mixtral 8x7BMistral AI · Open weights | 118.4 | 111.3 to 120.9 | Open weights (unrestricted) | $0.45 / $0.50 | 2023-12-11 |
80Yi-34B01.AI · Open weights | 117.3 | 108.9 to 120.6 | Open weights (restricted use) | $0.90 / $0.90 | 2023-11-02 |
81phi-3-mini 3.8BMicrosoft · Open weights | 117.3 | 109.6 to 120.8 | Open weights (unrestricted) | Not measured | 2024-04-23 |
82Stable Beluga 2Stability AI · Open weights | 117.0 | 109.6 to 120.1 | Open weights (non-commercial) | Not measured | 2023-07-20 |
83Llama 3.1-8BMeta · Open weights | 116.5 | 105.9 to 121.6 | Open weights (restricted use) | $0.10 / $0.10 | 2024-07-23 |
84Llama 3-8BMeta · Open weights | 116.4 | 107.5 to 119.3 | Open weights (restricted use) | $0.05 / $0.20 | 2024-04-18 |
85Gemma 3 4BGoogle · Open weights | 115.9 | 97.4 to 123.6 | Open weights (restricted use) | $0.04 / $0.08 | 2025-03-12 |
86Llama 2-70BMeta · Open weights | 113.7 | 104.6 to 117.2 | Open weights (restricted use) | $0.65 / $2.75 | 2023-07-18 |
87Qwen2.5-Coder (7B)Alibaba · Open weights | 113.0 | 102.9 to 119.2 | Open weights (unrestricted) | $0.035 / $0.075 | 2024-09-18 |
88Qwen-14BAlibaba · Open weights | 112.9 | 104.2 to 117.4 | Open weights (restricted use) | Not measured | 2023-09-24 |
89Mistral 7B v0.1Mistral AI · Open weights | 112.0 | 103.1 to 116.0 | Open weights (unrestricted) | $0.17 / $0.23 | 2023-09-27 |
90Falcon-180BTechnology Innovation Institute · Open weights | 112.0 | 102.2 to 119.5 | Open weights (restricted use) | Not measured | 2023-09-06 |
91Gemma 7BGoogle · Open weights | 111.8 | 103.1 to 116.7 | Open weights (restricted use) | $0.17 / $0.17 | 2024-02-21 |
92LLaMA-65BMeta · Open weights | 110.0 | 100.7 to 114.3 | Open weights (non-commercial) | Not measured | 2023-02-24 |
93Falcon 2 11BTechnology Innovation Institute · Open weights | 109.4 | 100.7 to 115.8 | Open weights (restricted use) | Not measured | 2024-05-09 |
94Mistral 7B v0.3Mistral AI · Open weights | 108.9 | 98.3 to 112.6 | Open weights (unrestricted) | $0.10 / $0.15 | 2024-05-27 |
95Phi-2Microsoft · Open weights | 107.7 | 86.3 to 113.5 | Open weights (unrestricted) | Not measured | 2023-12-12 |
96LLaMA-33BMeta · Open weights | 107.2 | 97.0 to 111.9 | Open weights (non-commercial) | Not measured | 2023-02-24 |
97Qwen-7BAlibaba · Open weights | 106.6 | 95.3 to 111.8 | Open weights (restricted use) | Not measured | 2023-09-28 |
98Llama 2-13BMeta · Open weights | 105.6 | 94.8 to 110.9 | Open weights (restricted use) | $0.10 / $0.50 | 2023-07-18 |
99StarCoder 2 15BHugging Face · Open weights | 104.7 | 91.3 to 111.7 | Open weights (restricted use) | $0.20 / $0.20 | 2024-02-20 |
100Yi 6B01.AI · Open weights | 104.5 | 94.0 to 110.3 | Open weights (restricted use) | $0.20 / $0.20 | 2023-11-22 |
How to read this leaderboard
Higher is better, and ranges that overlap mean a tie. Open weights is not the same as no strings attached. Some licences limit commercial use, size of company or purpose, so read the licence before you build on a model. The price column shows a hosted service's list price where we have one. Running a model yourself has hardware costs instead.
Questions people ask
What does open weights mean?
The trained model can be downloaded. Anyone with suitable hardware can run it, change it or host it. The training data and code may still be closed, which is why many people say open weights rather than open source.
Can I use an open-weights model for business?
Often yes, but not always. Licences differ. The access column shows how Epoch describes each model's terms. Always read the real licence on the maker's page before using a model commercially.
Why compare with models I have to pay for?
Open models can be run at your own cost or hosted by others. Check the overall leaderboard to see how they score next to closed models, and compare the cost of hosting yourself with the price of an API.
Why is a model I know missing?
A model appears only when an open source has published results for it. New models can take days or weeks to be measured, and models with no published score are left out rather than guessed.
Where does the data come from?
Scores come from Epoch AI's open benchmark data (licensed CC BY 4.0). We match models across sources, rank them and show them in charts, without changing the scores. Epoch AI does not endorse this site. Prices and context windows come from the open-source LiteLLM price list (MIT licence), which records each model maker's own price page. Prices change often, so confirm on the maker's site before you buy.
How often is this leaderboard updated?
We check our open data sources every day and show the date of the data at the top of the page. A model only moves when its source publishes new results, so a number can stay the same for weeks.
More leaderboards
- Overall intelligenceWhich AI model is the smartest overall?Leads: Claude Opus 5.5 (168.8)270 models ranked
- Human preferenceWhich AI do people prefer in blind tests?Leads: Gemini 4 Argon (1517)166 models ranked
- CodingWhich AI model is best for coding?Leads: Gemini 4 Argon (1552)166 models ranked
- Web developmentWhich AI model builds the best websites and apps?Leads: Claude Opus 5.5 (1807)85 models ranked
- Image understandingWhich AI model understands images best?Leads: Claude Fable 5 (1309)85 models ranked
- Cheapest strong modelsWhat is the cheapest AI model that is still good?Cheapest: gpt-oss-20b ($0.10)119 models ranked
- Best valueWhich AI models give the most intelligence for the money?Line starts at: gpt-oss-20b ($0.10)7 models ranked
- Context windowWhich AI model can read the most text at once?Leads: GPT-5.4 (1.1M tokens)193 models ranked
- GPQA DiamondWhich AI model scores highest on GPQA Diamond?Leads: Gemini 4 Argon (96.4%)183 models ranked
- Humanity's Last ExamWhich AI model scores highest on Humanity's Last Exam?Leads: GPT-6 Astra (54.8%)41 models ranked
- SWE-bench VerifiedWhich AI model scores highest on SWE-bench Verified?Leads: Claude Opus 4.7 (83.5%)32 models ranked
- Aider PolyglotWhich AI model scores highest on Aider Polyglot?Leads: GPT-5 (88.0%)46 models ranked
- Terminal-BenchWhich AI model scores highest on Terminal-Bench?Leads: GPT-5.3 Codex (64.7%)25 models ranked
- FrontierMath (Tiers 1-3)Which AI model scores highest on FrontierMath (Tiers 1-3)?Leads: GPT-6 Astra (93.7%)87 models ranked
- ARC-AGI-2Which AI model scores highest on ARC-AGI-2?Leads: GPT-6 Astra (95.0%)86 models ranked
- SimpleQA VerifiedWhich AI model scores highest on SimpleQA Verified?Leads: Gemini 4 Argon (76.0%)82 models ranked
- SimpleBenchWhich AI model scores highest on SimpleBench?Leads: Claude Fable 5 (81.9%)80 models ranked
- DeepSWEWhich AI model scores highest on DeepSWE?Leads: GPT-6 Astra (74.1%)26 models ranked
Where this data comes from
- Epoch AI: Benchmark data: Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks (CC BY 4.0). We match models across sources, rank them and show them in charts. Scores are not changed. Epoch AI does not endorse this site.
- LiteLLM model price list: Prices and context windows: LiteLLM model price list (github.com/BerriAI/litellm, MIT licence). We use the model maker's own price entry when there is one. Prices change often, so always confirm on the maker's site before you buy.
Newest data on this page: 26 Oct 2026. Prices are list prices and change often, so confirm on the maker's site before you buy. Sponsors never change a score or an order. How we calculate this · Report a mistake