Best LLM
What is the best LLM for reasoning?
GPT-6 Astra, Claude Fable 5.2, Gemini 4 Carbon and 2 others are practically level at the top in reasoning.
Among the ten best, GPT-6.1 Sol is the cheapest at $4 per million tokens.
The fastest among them is Claude Sonnet 5.5, at 103 tokens per second.
- Ranked models
- 76
- Tests
- 7
- Data as of
- Nov 10, 2026
Ranking: Reasoning
Place and value in the area, plus overall index, price and speed. Higher values are better.
| Rank | Model | Value | Overall index | Price (3:1) | Speed |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraTop group | 127.0 | 120.2 | $20 | 58 |
| 2 | Claude Fable 5.2Top group | 124.9 | 130.9 | $16 | n/a |
| 3 | GPT-6.1 SolUpper group | 122.9 | 118.0 | $4 | 44 |
| 4 | Gemini 4 CarbonUpper group | 122.7 | 127.0 | n/a | n/a |
| 5 | Claude Opus 5.5Upper group | 121.8 | 125.6 | $8 | 75 |
| 6 | Gemini 4 ArgonUpper group | 119.4 | 121.4 | $4 | 46 |
| 7 | Muse Wonder FullUpper group | 118.4 | 118.1 | n/a | n/a |
| 8 | Claude Sonnet 5.5Upper group | 118.3 | 121.8 | $4 | 103 |
| 9 | Gemini 3.7 FlashUpper group | 118.0 | 107.9 | n/a | 72 |
| 10 | GPT-5.6 SolUpper group | 116.7 | 109.1 | $8 | 75 |
| 11 | Claude Fable 5.1Upper group | 116.1 | 124.1 | $20 | 59 |
| 12 | GPT-6.1 TerraMiddle group | 114.7 | 106.5 | $2 | 62 |
| 13 | Muse WonderMiddle group | 114.6 | 113.3 | $8 | n/a |
| 14 | Kimi K3.1 ProMiddle group | 114.4 | 112.6 | $4 | 56 |
| 15 | GPT-6 SolMiddle group | 114.3 | 109.1 | $4 | 82 |
| 16 | Gemini 3.8 FlashMiddle group | 114.3 | 106.8 | $1.50 | 81 |
| 17 | Muse Spark 1.3Middle group | 110.2 | 107.3 | $2 | 102 |
| 18 | Muse Ember 1.0Middle group | 108.3 | 110.4 | $0.20 | n/a |
| 19 | Claude Opus 5Middle group | 107.7 | 117.4 | $10 | 71 |
| 20 | GPT-5.6 TerraMiddle group | 105.7 | 100.8 | $4.50 | 72 |
Show all 76 ranked models
| Rank | Model | Value | Overall index | Price (3:1) | Speed |
|---|---|---|---|---|---|
| 21 | Claude Haiku 5.5Middle group | 105.4 | 102.8 | $0.20 | 164 |
| 22 | Gemini 3.1 Pro PreviewMiddle group | 104.2 | 91.4 | $4.50 | 96 |
| 23 | Gemini 3.6 FlashMiddle group | 103.4 | 97.6 | $1.50 | 110 |
| 24 | Claude Fable 5Middle group | 103.3 | 112.8 | $20 | 50 |
| 25 | Qwen 3.8 Max 0902Middle group | 103.2 | 100.8 | $2.48 | 36 |
| 26 | Grok 4.7Middle group | 102.2 | 103.3 | $3 | 47 |
| 27 | MiniMax M3.1 FlashMiddle group | 101.5 | 100.0 | $0.26 | n/a |
| 28 | MiMo V2.6 ProMiddle group | 101.3 | 100.4 | $0.54 | 32 |
| 29 | DeepSeek V4.1 Flash 1017Middle group | 100.8 | 106.7 | $0.26 | 126 |
| 30 | Grok 4.6Middle group | 100.5 | 102.1 | $3 | 58 |
| 31 | GPT-5.6 LunaMiddle group | 100.0 | 93.8 | $0.45 | 123 |
| 32 | Grok 4.5Middle group | 99.7 | 94.8 | $3 | 52 |
| 33 | GPT-5.5Middle group | 99.4 | 101.5 | $11.25 | 87 |
| 34 | Mistral Large 4Middle group | 99.1 | 100.8 | $1.05 | n/a |
| 35 | GPT-5.4 ProMiddle group | 98.8 | n/a | $67.50 | n/a |
| 36 | GPT-6 LunaMiddle group | 98.0 | 95.8 | $0.20 | 77 |
| 37 | Muse Spark OpenMiddle group | 97.7 | 97.5 | n/a | n/a |
| 38 | Kimi K3Middle group | 97.2 | 101.1 | $6 | 61 |
| 39 | Ling 3.1 Flash 1019Middle group | 97.1 | n/a | n/a | n/a |
| 40 | Step 5 PreviewMiddle group | 96.9 | 97.1 | $1.43 | 65 |
| 41 | GLM-5.3Middle group | 95.9 | 100.6 | $2.15 | 81 |
| 42 | Gemini 3.5 FlashMiddle group | 95.5 | 97.8 | n/a | 142 |
| 43 | MiMo V2.6 FlashMiddle group | 92.7 | 97.3 | $0.18 | 38 |
| 44 | Agnes 2.5 Pro ThetaMiddle group | 92.4 | n/a | $0.15 | n/a |
| 45 | Claude Sonnet 5Middle group | 90.4 | 92.1 | $4 | 77 |
| 46 | MiniMax M3Middle group | 89.5 | 83.0 | $0.52 | 82 |
| 47 | Claude Opus 4.7Middle group | 89.2 | 95.1 | $10 | 57 |
| 48 | GPT-5.4Middle group | 88.0 | 86.5 | $5.63 | 58 |
| 49 | GLM-5.2Middle group | 87.3 | 89.0 | $2.15 | 73 |
| 50 | Claude Opus 4.8Middle group | 87.3 | 99.5 | $10 | 71 |
| 51 | Reflection BeamMiddle group | 87.1 | 82.4 | n/a | n/a |
| 52 | GLM-5.3-FlashMiddle group | 87.0 | 94.3 | $0.24 | 101 |
| 53 | DeepSeek-V4-ProLower group | 84.3 | 87.2 | $0.99 | 44 |
| 54 | Claude Opus 4.6Lower group | 84.3 | 84.2 | $10 | 36 |
| 55 | GPT-5 ProLower group | 82.5 | n/a | $41.25 | n/a |
| 56 | Qwen 3.8 27BLower group | 81.1 | 82.8 | $0.66 | 66 |
| 57 | InklingLower group | 81.1 | 72.1 | $1.76 | n/a |
| 58 | K2 Horizon 375B-A23BLower group | 77.1 | n/a | n/a | n/a |
| 59 | Inkling SmallLower group | 77.0 | 68.7 | $0.52 | 61 |
| 60 | Nemotron 3.5 Super VL 120B A12BLower group | 73.6 | 68.1 | $0.24 | n/a |
| 61 | Gemini 3.5 Flash-LiteLower group | 72.2 | 73.7 | $0.85 | 145 |
| 62 | o3Lower group | 71.8 | 74.2 | $3.50 | 73 |
| 63 | GPT-5Lower group | 70.8 | 77.5 | $3.44 | 83 |
| 64 | Nemotron 3 UltraLower group | 70.6 | 66.8 | $0.93 | 92 |
| 65 | Muse Glimmer 30BLower group | 69.1 | n/a | $0.52 | 124 |
| 66 | Gemini 3.1 Flash-LiteLower group | 60.2 | 55.9 | $0.56 | 125 |
| 67 | GPT OSS 120BLower group | 52.1 | 43.4 | $0.20 | 92 |
| 68 | Claude Haiku 4.5Lower group | 50.1 | 54.1 | $2 | 84 |
| 69 | Nemotron 3.5 LightningLower group | 39.3 | 31.1 | $0.085 | 180 |
| 70 | GPT-4.1Lower group | 39.0 | 43.6 | $3.50 | 81 |
| 71 | North Mini CodeLower group | 33.8 | n/a | n/a | n/a |
| 72 | Command A PlusLower group | 32.2 | 30.7 | n/a | n/a |
| 73 | Mistral Medium 3.5Lower group | 28.7 | 46.3 | $3 | 118 |
| 74 | Granite 4.2 8BLower group | 21.4 | n/a | n/a | n/a |
| 75 | Granite 4.2 3BLower group | 17.2 | n/a | n/a | n/a |
| 76 | Mistral Small 4Lower group | 15.2 | 23.3 | $0.26 | 112 |
Price in dollars per million tokens, mixed from 3 parts input and 1 part output. Speed in tokens per second.
GPT-6 AstraRank 1127.0
Claude Fable 5.2Rank 2124.9
GPT-6.1 SolRank 3122.9
Gemini 4 CarbonRank 4122.7
Claude Opus 5.5Rank 5121.8
Gemini 4 ArgonRank 6119.4
Muse Wonder FullRank 7118.4
Claude Sonnet 5.5Rank 8118.3
Gemini 3.7 FlashRank 9118.0
GPT-5.6 SolRank 10116.7
Claude Fable 5.1Rank 11116.1
GPT-6.1 TerraRank 12114.7
Muse WonderRank 13114.6
Kimi K3.1 ProRank 14114.4
GPT-6 SolRank 15114.3
Gemini 3.8 FlashRank 16114.3
Muse Spark 1.3Rank 17110.2
Muse Ember 1.0Rank 18108.3
Claude Opus 5Rank 19107.7
GPT-5.6 TerraRank 20105.7
Which tests count here?
7 tests by independent evaluators feed this area. Each test counts with the result of the model in question.
GPQA Diamond
90 models tested · Epoch AI benchmark run dataset snapshot, Vals AI
CritPt v2 (Hard Subset)
57 models tested · Artificial Analysis
GDP.pdf
57 models tested · Artificial Analysis
Chess Puzzles
22 models tested · Epoch AI benchmark run dataset snapshot
MultiNRC
12 models tested · Scale AI
Humanity’s Last Exam
7 models tested · Scale AI
EnigmaEval
5 models tested · Scale AI
Popular comparisons among the best
Other tasks
All LLMs comparedHow the overall index worksCompare several models