AI Model Leaderboard
The only board that ranks LLMs on all three axes that matter to a builder: how smart they are (intelligence), how widely used they are (real OpenRouter traffic), and your value — intelligence per dollar.
Updated September 14, 2026
Ranked by the Artificial Analysis Intelligence Index. Value = intelligence per blended $/1M token. Popularity shown where the model has OpenRouter traffic.
| # | Model | |||
|---|---|---|---|---|
| 1 | Claude Fable 5.1 | 53.4 | #48 1% | 1.78 |
| 2 | GPT-6 Astra | 52.8 | #46 188% | 1.76 |
| 3 | Claude Opus 5 | 50.7 | #14 49% | 3.38 |
| 4 | Claude Fable 5 | 49.7 | #53 100% | 1.66 |
| 5 | Muse Spark 1.3 | 48.2 | — | 17.53 |
| 6 | GPT-5.6 Sol | 47.1 | #15 33% | 3.93 |
| 7 | GLM-5.3 | 44.9 | #13 40% | 15.48 |
| 8 | Grok 4.6 | 44.4 | #34 5% | 11.10 |
| 9 | GPT-5.6 Terra | 42.3 | #27 21% | 6.04 |
| 10 | GLM-5.3-Flash | 41.9 | #4 18% | 128.92 |
| 11 | Gemini 3.8 Flash | 41.2 | #22 70% | 18.31 |
| 12 | Claude Opus 4.7 | 40.7 | #55 — | 2.71 |
| 13 | Qwen3.8 Max | 40.3 | #41 0% | 10.07 |
| 14 | Qwen3.8 2.4T A95B | 40.0 | — | 10.00 |
| 15 | Qwen3.8-Flash-Next | 39.9 | — | 128.71 |
| 16 | Muse Spark 1.2 | 39.8 | — | 14.47 |
| 17 | DeepSeek V4.1 Flash | 39.5 | #19 — | 52.67 |
| 18 | Gemini 3.7 Flash | 39.4 | #12 64% | 17.51 |
| 19 | Grok 4.5 | 39.1 | #76 — | 9.78 |
| 20 | GPT-5.4 | 39.0 | #68 — | 4.46 |
| 21 | GPT-5.5 | 38.6 | #54 — | 2.21 |
| 22 | Claude Sonnet 5 | 38.4 | #18 10% | 6.40 |
| 23 | GPT-5.6 Luna | 37.5 | #2 10% | 53.57 |
| 24 | DeepSeek V4 Pro 0813 | 36.3 | — | 13.75 |
| 25 | Agnes 3.0 Flash | 35.5 | — | 355.00 |
| 26 | DeepSeek V4 Flash Vision | 35.0 | — | 39.77 |
| 27 | Kimi K3 | 34.5 | #16 36% | 3.83 |
| 28 | Gemini 3.6 Flash | 34.3 | #29 100% | 15.24 |
| 29 | Muse Spark 1.1 | 34.3 | — | 12.47 |
| 30 | Qwen3.8 27B | 33.9 | #56 83% | 19.37 |
| 31 | Gemini 3.5 Flash | 33.6 | #71 — | 6.40 |
| 32 | Motif 3 | 33.6 | — | — |
| 33 | GPT-5.3 Codex | 32.5 | — | 4.13 |
| 34 | Claude Opus 4.6 | 31.9 | #69 — | 2.13 |
| 35 | Kimi K2.6 | 31.3 | #66 100% | 12.65 |
| 36 | DeepSeek V4 Pro | 30.9 | #11 7% | 47.54 |
| 37 | K2 Horizon 375B A23B | 30.8 | — | — |
| 38 | Claude Sonnet 4.6 | 30.5 | #24 64% | 3.39 |
| 39 | Apodex 1.1 | 30.4 | — | 18.42 |
| 40 | GPT-5.2 | 30.4 | — | 3.86 |
| 41 | Gemini 3.1 Pro Preview | 30.4 | #52 42% | 4.34 |
| 42 | Qwen3.7 Max | 29.9 | #75 — | 5.98 |
| 43 | MiniMax-M3 | 29.6 | #9 81% | 39.47 |
| 44 | Claude Opus 4.5 | 29.1 | — | 1.94 |
| 45 | MiMo-V2-Pro | 28.6 | — | — |
| 46 | GPT-5.2 Codex | 28.5 | — | 3.62 |
| 47 | Qwen3.6 Max Preview | 28.4 | — | 6.24 |
| 48 | Solar Pro 4 | 28.2 | #20 13% | 37.60 |
| 49 | Nex-N2-Pro | 28.2 | — | 18.80 |
| 50 | Gemini 3 Pro Preview | 28.0 | — | 4.00 |
The board is a superset of the calculator catalog: brand-new models appear here with quality & popularity before the pricing catalog catches up. Models with a calculator match link to a head-to-head comparison.
Data: Artificial Analysis · OpenRouter
Know your margin before you build
Pick a model above, then model your real cost and price in the Calcaas calculator — free.