LiveBench long agentic leaderboard
- Claude Opus 5.5 leads
- 85.0%
- models measured
- 43
- lead over #2
- +2.6 pts
- median
- 40.4%
Leader on LiveBench long agentic · read October 22
On October 22, 2026, 43 of our catalog's models had a LiveBench long agentic result, and Anthropic Claude Opus 5.5 led with 85.0%, ahead of OpenAI GPT-6 Astra at 82.4%.
Long-running work: applications extended over several turns of changing requirements, an existing codebase operated over simulated weeks or months, and extended coding and optimisation tasks, all graded deterministically. Each score is the model's mean task score for the category, 0 to 100, from a single run of the current release, shown as a share.
Role in the index
LiveBench long agentic carries 30% of the Agentic Index. A model missing it is scored on the other benchmarks of that index. Read the methodology
| # | Model | Result | Cost per task | Output tokens per task | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 85.0% | $38.0088 | 463,030 | 88.8 | $20.00 | |
| 2 | GPT-6 AstraOpenAI | 82.4% | $38.3789 | 210,429 | 85.1 | $50.00 | |
| 3 | Claude Fable 5.1Anthropic | 80.8% | $51.5728 | 313,679 | 84.2 | $50.00 | |
| 4 | Claude Sonnet 5.5Anthropic | 78.8% | $35.0115 | 773,762 | 87.3 | $10.00 | |
| 5 | GPT-6.1 SolOpenAI | 73.0% | $6.4877 | 194,893 | 82.4 | $10.00 | |
| 6 | Gemini 4 ArgonGoogle | 69.9% | $13.8419 | 327,753 | 81.0 | $10.00 | |
| 7 | GPT-6 SolOpenAI | 66.9% | $8.6215 | 212,220 | 76.0 | $10.00 | |
| 8 | Claude Opus 5Anthropic | 62.1% | $48.7532 | 384,099 | 83.0 | $25.00 | |
| 9 | Claude Haiku 5.5Anthropic | 57.6% | $10.3619 | 1.0M | 70.7 | $0.50 | |
| 10 | Claude Fable 5Anthropic | 53.2% | $59.7366 | 265,052 | 79.3 | $50.00 | |
| 11 | DeepSeek V4.1 Flash 1017DeepSeek | 50.9% | $1.4582 | 417,570 | 70.3 | $0.60 | |
| 12 | Grok 4.7xAI | 50.1% | $18.9333 | 283,495 | 66.7 | $6.00 | |
| 13 | GPT-5.6 SolOpenAI | 48.5% | $15.9615 | 164,160 | 73.5 | $20.00 | |
| 14 | GPT-5.6 TerraOpenAI | 48.4% | $11.7569 | 253,005 | 63.9 | $12.00 | |
| 15 | Claude Opus 4.8Anthropic | 48.3% | $51.9696 | 522,691 | 65.0 | $25.00 | |
| 16 | GPT-6 LunaOpenAI | 48.1% | $0.9676 | 688,782 | 57.3 | $0.50 | |
| 17 | MiMo-V2.6-ProXiaomi | 47.5% | $1.3886 | 417,649 | 65.4 | $0.87 | |
| 18 | GPT-5.5OpenAI | 44.6% | $13.8515 | 95,488 | 71.3 | $30.00 | |
| 19 | Kimi K3Moonshot AI | 42.8% | $17.9905 | 272,271 | 62.3 | $15.00 | |
| 20 | Qwen3.8 Max (0902)Qwen | 42.3% | $14.93 | 386,822 | 62.4 | $6.00 | |
| 21 | Grok 4.6xAI | 40.7% | $10.6258 | 128,260 | 65.7 | $6.00 | |
| 22 | Muse Spark 1.3Meta | 40.4% | $8.839 | 165,448 | 57.4 | $4.25 | |
| 23 | DeepSeek V4.1 FlashDeepSeek | 40.1% | $1.7415 | 451,375 | 62.5 | $0.84 | |
| 24 | Claude Sonnet 5Anthropic | 39.0% | $33.2111 | 554,955 | 57.7 | $10.00 | |
| 25 | DeepSeek V4 Flash Vision ExpDeepSeek | 39.0% | $1.0407 | 357,320 | Not benchmarked | $1.32 | |
| 26 | Qwen3.8 FlashQwen | 37.1% | $1.872 | 459,461 | Not benchmarked | $0.47 | |
| 27 | GLM 5.3Z.ai | 35.9% | $18.3819 | 443,888 | 70.6 | $4.40 | |
| 28 | GPT-5.6 LunaOpenAI | 30.4% | $1.1012 | 234,057 | 58.0 | $1.20 | |
| 29 | Gemini 3.8 FlashGoogle | 30.1% | $7.039 | 284,589 | 61.6 | $3.75 | |
| 30 | GLM 5.3 FlashZ.ai | 28.4% | $1.6288 | 269,870 | 57.7 | $0.50 | |
| 31 | Gemini 3.7 FlashGoogle | 28.3% | $5.5585 | 205,813 | 58.0 | $3.75 | |
| 32 | Qwen3.8 27BQwen | 27.9% | $7.1841 | 366,727 | 74.2 | $2.55 | |
| 33 | Grok 4.5xAI | 27.9% | $4.4983 | 80,606 | 58.6 | $6.00 | |
| 34 | Step 5 PreviewStepFun | 27.5% | $10.3756 | 493,230 | 61.1 | $2.70 | |
| 35 | Claude Sonnet 4.6Anthropic | 23.1% | $20.4196 | 310,212 | 57.3 | $15.00 | |
| 36 | Muse Spark 1.2Meta | 21.3% | $3.9488 | 122,638 | 55.7 | $4.25 | |
| 37 | Muse Spark OpenMeta | 18.9% | $4.8336 | 133,814 | 47.4 | $4.25 | |
| 38 | Nemotron 3.5 Super VLNVIDIA | 15.8% | $2.0477 | 220,348 | Not benchmarked | $0.50 | |
| 39 | T InklingThinking Machines | 14.1% | $12.4056 | 168,123 | 31.5 | $4.05 | |
| 40 | Gemma 4 31BGoogle | 13.3% | $0.9887 | 42,056 | 39.0 | $0.40 | |
| 41 | MiniMax M3MiniMax | 12.4% | $4.1953 | 254,681 | 39.2 | $1.20 | |
| 42 | Mistral Large 4Mistral AI | 11.8% | $9.5557 | 508,413 | 56.4 | $4.18 | |
| 43 | Claude Haiku 4.5Anthropic | 11.1% | $2.5485 | 77,255 | 35.6 | $5.00 |
Only models in our catalog are listed, and each appears once with its best published run. 43 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Cost per task is what LiveBench reports the run spent on the category, divided by the number of tasks in it, in US dollars; LiveBench defines it the same way. Every other column is a figure the board publishes for the same run.
| Model | Result | Finance & Compliance Apps | Operations & Commerce Apps | Long-Run Operations | Systems Engineering | Performance Optimisation |
|---|---|---|---|---|---|---|
| Anthropic Claude Opus 5.5 | 85.0% | 100.0% | 87.5% | 45.4% | 98.1% | 87.6% |
| OpenAI GPT-6 Astra | 82.4% | 100.0% | 100.0% | 29.1% | 98.3% | 75.1% |
| Anthropic Claude Fable 5.1 | 80.8% | 95.8% | 81.2% | 47.8% | 92.8% | 81.0% |
| Anthropic Claude Sonnet 5.5 | 78.8% | 100.0% | 81.2% | 36.1% | 84.9% | 82.5% |
| OpenAI GPT-6.1 Sol | 73.0% | 66.7% | 93.8% | 20.8% | 94.6% | 81.5% |
| Google Gemini 4 Argon | 69.9% | 91.7% | 66.7% | 19.4% | 97.6% | 68.4% |
| OpenAI GPT-6 Sol | 66.9% | 83.3% | 54.2% | 38.9% | 85.2% | 70.5% |
| Anthropic Claude Opus 5 | 62.1% | 62.5% | 60.4% | 23.6% | 95.2% | 67.3% |
| Anthropic Claude Haiku 5.5 | 57.6% | 70.8% | 43.8% | 33.3% | 62.8% | 72.7% |
| Anthropic Claude Fable 5 | 53.2% | 58.4% | 47.9% | 1.4% | 88.3% | 65.9% |
| DeepSeek V4.1 Flash 1017 | 50.9% | 41.7% | 37.5% | 34.7% | 82.2% | 62.0% |
| SpaceXAI Grok 4.7 | 50.1% | 50.0% | 50.0% | 5.6% | 79.3% | 61.7% |
| OpenAI GPT-5.6 Sol | 48.5% | 54.2% | 41.7% | 16.7% | 87.8% | 44.2% |
| OpenAI GPT-5.6 Terra | 48.4% | 54.1% | 27.1% | 11.1% | 79.1% | 68.7% |
| Anthropic Claude Opus 4.8 | 48.3% | 54.1% | 31.2% | 11.1% | 73.9% | 68.3% |
| OpenAI GPT-6 Luna | 48.1% | 45.9% | 43.8% | 15.3% | 67.8% | 64.4% |
| Xiaomi MiMo-V2.6-Pro | 47.5% | 29.2% | 43.8% | 13.9% | 81.5% | 69.3% |
| OpenAI GPT-5.5 | 44.6% | 37.5% | 16.7% | 22.2% | 81.4% | 68.8% |
| MoonshotAI Kimi K3 | 42.8% | 37.5% | 8.3% | 6.9% | 86.4% | 76.8% |
| Qwen3.8 Max (0902) | 42.3% | 29.1% | 22.9% | 12.5% | 81.4% | 67.8% |
| SpaceXAI Grok 4.6 | 40.7% | 37.5% | 35.4% | 5.6% | 80.4% | 45.6% |
| Meta Muse Spark 1.3 | 40.4% | 29.2% | 33.3% | 6.9% | 69.9% | 61.5% |
| DeepSeek V4.1 Flash | 40.1% | 20.8% | 41.7% | 1.4% | 87.2% | 51.6% |
| Anthropic Claude Sonnet 5 | 39.0% | 16.6% | 39.6% | 5.6% | 70.1% | 62.4% |
| DeepSeek V4 Flash Vision Exp | 39.0% | 37.5% | 33.3% | 0.0% | 83.5% | 42.0% |
| Qwen3.8 Flash | 37.1% | 12.5% | 45.8% | 18.1% | 65.3% | 46.2% |
| Z.ai GLM 5.3 | 35.9% | 37.5% | 27.1% | 0.0% | 65.0% | 48.3% |
| OpenAI GPT-5.6 Luna | 30.4% | 12.5% | 12.5% | 0.0% | 69.5% | 59.6% |
| Google Gemini 3.8 Flash | 30.1% | 16.7% | 8.3% | 11.1% | 78.2% | 43.5% |
| Z.ai GLM 5.3 Flash | 28.4% | 8.3% | 20.9% | 0.0% | 79.0% | 39.5% |
| Google Gemini 3.7 Flash | 28.3% | 0.0% | 8.3% | 0.0% | 71.2% | 65.8% |
| Qwen3.8 27B | 27.9% | 8.3% | 25.0% | 0.0% | 67.3% | 41.6% |
| SpaceXAI Grok 4.5 | 27.9% | 12.5% | 0.0% | 0.0% | 67.6% | 62.4% |
| StepFun Step 5 Preview | 27.5% | 4.2% | 16.7% | 5.6% | 75.2% | 42.5% |
| Anthropic Claude Sonnet 4.6 | 23.1% | 0.0% | 10.4% | 0.0% | 68.2% | 42.6% |
| Meta Muse Spark 1.2 | 21.3% | 4.2% | 4.2% | 0.0% | 67.9% | 36.7% |
| Meta Muse Spark Open | 18.9% | 12.5% | 4.2% | 0.0% | 44.2% | 35.3% |
| NVIDIA Nemotron 3.5 Super VL | 15.8% | 4.2% | 4.2% | 0.0% | 35.7% | 35.8% |
| Thinking Machines Inkling | 14.1% | 0.0% | 0.0% | 0.0% | 38.0% | 34.9% |
| Google Gemma 4 31B | 13.3% | 4.2% | 8.3% | 0.0% | 19.6% | 32.7% |
| MiniMax M3 | 12.4% | 0.0% | 0.0% | 0.0% | 29.4% | 34.0% |
| Mistral Large 4 | 11.8% | 0.0% | 0.0% | 0.0% | 19.6% | 38.2% |
| Anthropic Claude Haiku 4.5 | 11.1% | 0.0% | 0.0% | 0.0% | 13.8% | 39.6% |
The parts as LiveBench publishes them, for the same run as the result.