v1.3 Preview
Best AI models for knowledge work
Compare models for client research, coding and professional deliverables so you can choose what work to delegate. This index combines 7 benchmarks; it does not measure hours saved or the share of a project a model can complete.
Updated: Data as of:
Rankings
Agent Index (0–100)- 01
Claude Fable 5.289.6 - 02Claude Opus 5.589.2
- 03
GPT-6 Astra86.2 - 04Claude Sonnet 5.583.5
- 05Claude Fable 5.183.4
- 06GPT-6.1 Sol80.6
- 07GPT-6.1 Terra69.2
- 08
Gemini 4 Argon65.9 - 09
Kimi K3.1 Pro65.4 - 10
Muse Wonder 1.057.3 - 11Claude Haiku 5.546.2
- 12Muse Spark 1.341.5
- 13
DeepSeek V4.1 Flash 101739.6 - 14
MiMo-V2.6-Pro39.1 - 15
MiniMax M3.1 Flash36.8
- 16
Grok 4.736.7 - 17GPT-5.6 Terra35.2
- 18Kimi K334.4
- 19
GLM-5.327.3 - 20GPT-6 Luna24.4
- 21GLM-5.3-Flash22.6
- 22
Step 5 Preview16.4 - 23Gemini 3.8 Flash10.4
- 24
Qwen3.8 27B8.1 - 25Muse Spark Open6.1
- 26
Nemotron 3.5 Super VL1.6 - 27
Inkling0.0
Score breakdown
Each component is normalized to 0–100 using fixed v1 reference points. Raw benchmark scores appear below.
| Model | HLE-Diamond10% weight | GDP.pdf10% weight | AA-Briefcase20% weight | LiveBench Agentic20% weight | LiveBench Coding5% weight | Terminal-Bench 4.015% weight | LiveBench Browser Use20% weight | Agent IndexScore / 100 |
|---|---|---|---|---|---|---|---|---|
| Claude Fable 5.2Anthropic | 100.080.2% raw | 71.326.8% raw | 95.31,785.32 Elo raw | 100.082.9% raw | 51.879.2% raw | 89.861.62% raw | 86.786.9% raw | 89.6 |
| Claude Opus 5.5Anthropic | 100.067.9% raw | 67.526.2% raw | 97.91,801 Elo raw | 91.772.8% raw | 72.487.1% raw | 86.159.6% raw | 89.888.4% raw | 89.2 |
| GPT-6 AstraOpenAI | 91.863.8% raw | 100.032.2% raw | 59.91,569 Elo raw | 96.375.6% raw | 73.487.5% raw | 86.159.6% raw | 95.991.3% raw | 86.2 |
| Claude Sonnet 5.5Anthropic | 70.653.3% raw | 65.025.8% raw | 100.01,814 Elo raw | 76.463.4% raw | 46.677.2% raw | 93.563.64% raw | 91.789.3% raw | 83.5 |
| Claude Fable 5.1Anthropic | 84.360.1% raw | 67.526.2% raw | 76.41,670 Elo raw | 94.074.2% raw | 79.289.7% raw | 77.855.05% raw | 92.789.8% raw | 83.4 |
| GPT-6.1 SolOpenAI | 81.758.8% raw | 100.032% raw | 57.61,555 Elo raw | 80.065.6% raw | 59.482.1% raw | 79.656.06% raw | 100.093.3% raw | 80.6 |
| GPT-6.1 TerraOpenAI | 58.647.3% raw | 82.528.6% raw | 32.61,402 Elo raw | 64.756.2% raw | 77.989.2% raw | 78.855.6% raw | 99.693.1% raw | 69.2 |
| Gemini 4 ArgonGoogle | 77.556.7% raw | 40.021.8% raw | 46.01,484 Elo raw | 68.458.5% raw | 68.885.7% raw | 81.557.07% raw | 78.282.8% raw | 65.9 |
| Kimi K3.1 ProMoonshot AI | 38.437.3% raw | 100.031.4% raw | 77.41,675.78 Elo raw | 36.238.7% raw | 22.167.8% raw | 100.067.17% raw | 63.975.9% raw | 65.4 |
| Muse Wonder 1.0Meta | 41.939% raw | 95.030.6% raw | 74.61,659 Elo raw | 48.446.2% raw | 40.674.9% raw | 72.252% raw | 30.960% raw | 57.3 |
| Claude Haiku 5.5Anthropic | 22.929.6% raw | 33.820.8% raw | 61.01,576 Elo raw | 51.548.1% raw | 47.777.6% raw | 37.032.8% raw | 50.669.5% raw | 46.2 |
| Muse Spark 1.3Meta | 31.233.7% raw | 70.026.6% raw | 61.91,581 Elo raw | 51.147.9% raw | 62.083.1% raw | 38.033.33% raw | 0.030.9% raw | 41.5 |
| DeepSeek V4.1 Flash 1017DeepSeek | 14.325.3% raw | 11.217.2% raw | 60.11,570.42 Elo raw | 36.338.8% raw | 54.980.4% raw | 38.933.84% raw | 45.967.2% raw | 39.6 |
| MiMo-V2.6-ProXiaomi | 0.015.2% raw | 23.719.2% raw | 50.61,512 Elo raw | 32.436.4% raw | 13.564.5% raw | 40.734.8% raw | 66.877.3% raw | 39.1 |
| MiniMax M3.1 FlashMiniMax | 19.928.1% raw | 33.820.8% raw | 44.51,474.82 Elo raw | 15.025.7% raw | 24.768.8% raw | 24.125.76% raw | 73.780.6% raw | 36.8 |
| Grok 4.7xAI | 10.923.6% raw | 48.823.2% raw | 71.21,638 Elo raw | 46.945.3% raw | 35.973.1% raw | 24.125.76% raw | 8.749.3% raw | 36.7 |
| GPT-5.6 TerraOpenAI | 21.328.8% raw | 57.524.6% raw | 21.31,333 Elo raw | 15.325.9% raw | 64.884.2% raw | 41.835.4% raw | 52.370.3% raw | 35.2 |
| Kimi K3Moonshot AI | 17.526.9% raw | 41.322% raw | 48.41,498.42 Elo raw | 31.836% raw | 0.053.1% raw | 0.012.63% raw | 62.475.2% raw | 34.4 |
| GLM-5.3Z.ai | 0.017.9% raw | 0.011.2% raw | 49.41,504.57 Elo raw | 31.635.9% raw2.0 / 2026-10-14 | 61.282.8% raw2.0 / 2026-10-14 | 53.741.92% raw | 0.00% raw2.0 / 2026-10-14 | 27.3 |
| GPT-6 LunaOpenAI | 8.022.2% raw | 46.322.8% raw | 21.91,337 Elo raw | 13.424.7% raw | 29.970.8% raw | 0.012.6% raw | 52.370.3% raw | 24.4 |
| GLM-5.3-FlashZ.ai | 0.018.2% raw | 0.015.4% raw | 40.51,450.23 Elo raw | 6.820.7% raw | 30.270.9% raw | 37.032.83% raw | 30.559.8% raw | 22.6 |
| Step 5 PreviewStepFun | 12.324.3% raw | 0.014.8% raw | 36.31,425 Elo raw | 4.719.4% raw | 26.369.4% raw | 37.933.3% raw | 0.024.8% raw | 16.4 |
| Gemini 3.8 FlashGoogle | 30.033.1% raw | 46.322.8% raw | 0.01,203 Elo raw | 0.016.5% raw | 16.765.7% raw | 13.019.7% raw | 0.045.1% raw | 10.4 |
| Qwen3.8 27BQwen | 0.015.3% raw | 6.216.4% raw | 31.11,393 Elo raw | 0.015.9% raw | 26.069.3% raw | 0.05.6% raw | 0.024.2% raw | 8.1 |
| Muse Spark OpenMeta | 16.326.3% raw | 20.018.6% raw | 12.41,279 Elo raw | 0.012.1% raw | 0.055.1% raw | 0.08.1% raw | 0.026.2% raw | 6.1 |
| Nemotron 3.5 Super VLNVIDIA | 0.08.8% raw | 0.08% raw | 0.0999 Elo raw | 0.011.7% raw2.0 / 2026-10-25 | 31.271.3% raw2.0 / 2026-10-25 | 0.00% raw | 0.021.6% raw2.0 / 2026-10-25 | 1.6 |
| InklingThinking Machines | 0.010.6% raw | 0.012.8% raw | 0.0833 Elo raw | 0.09.8% raw | 0.048.7% raw | 0.01% raw | 0.012% raw | 0.0 |
Sources
Published evaluations behind each component of Agent Index. View the original sources ↗
Models awaiting complete coverage 70
- DeepSeek V4 Pro 08134/7 results
- DeepSeek V4.1 Flash4/7 results
- Gemini 3.1 Pro4/7 results
- Gemini 4 Carbon4/7 results
- gpt-oss-120b4/7 results
- gpt-oss-20b4/7 results
- MiniMax M34/7 results
- Qwen 3.5 9B4/7 results
- Qwen3.5-397B-A17B4/7 results
- Claude Opus 4.83/7 results
- Claude Opus 53/7 results
- Claude Sonnet 4.53/7 results
- Claude Sonnet 4.63/7 results
- Claude Sonnet 53/7 results
- DeepSeek R1 (Jan 2025)3/7 results
- DeepSeek V4 Flash3/7 results
- DeepSeek V4 Flash 04203/7 results
- DeepSeek V4 Pro 04243/7 results
- GLM-5.13/7 results
- GLM-5.23/7 results
- GPT-5.4 mini3/7 results
- GPT-5.53/7 results
- GPT-5.6 Luna3/7 results
- GPT-5.6 Sol3/7 results
- Kimi K2.63/7 results
- MiniMax M2.73/7 results
- Mistral Large 43/7 results
- Muse Spark 1.13/7 results
- Muse Spark 1.23/7 results
- Qwen 3.8 Max3/7 results
- Qwen3-14B3/7 results
- Claude Fable 51/7 results
- Claude Opus 4.71/7 results
- Muse Spark1/7 results
- Claude Mythos 50/7 results
- Claude Mythos 5.10/7 results
- Claude Mythos 5.20/7 results
- Claude Mythos Preview0/7 results
- Claude Opus 4.50/7 results
- Claude Opus 4.60/7 results
- DeepSeek V3.20/7 results
- DeepSeek V4 Pro0/7 results
- DeepSeekCoder0/7 results
- Gemini 3 Flash0/7 results
- Gemini 3 Pro0/7 results
- GLM-50/7 results
- GPT-4 Turbo0/7 results
- GPT-4.10/7 results
- GPT-4.1 mini0/7 results
- GPT-4o0/7 results
- GPT-5.20/7 results
- GPT-5.40/7 results
- GPT-5.4 Pro0/7 results
- GPT-5.5 Pro0/7 results
- GPT-5.60/7 results
- GPT-6 Cyber0/7 results
- Grok 4 Heavy0/7 results
- Inkling 2 Small0/7 results
- Inkling-Small0/7 results
- Kimi K2 Thinking0/7 results
- Kimi K2.50/7 results
- Ling-3.2-flash0/7 results
- LLaMA3-70B0/7 results
- Qwen 3.5 35B-A3B0/7 results
- Qwen2.5-32B0/7 results
- Qwen3.6 Plus0/7 results
- Ripple-65B Preview0/7 results
- Seed2.0 Pro0/7 results
- Step 3.7 Flash0/7 results
- SWE-2 Flash0/7 results