LiveBench coding leaderboard
- Claude Fable 5 leads
- 98.3%
- models measured
- 43
- lead over #2
- +0.6 pts
- median
- 84.5%
Leader on LiveBench coding · read October 21
# LiveBench coding, read October 21
leaders = {
"Claude Fable 5": 98.3%, "Claude Fable 5.1": 97.7%, "Claude Opus 5.5": 96.3%,
}
On October 21, 2026, 43 of our catalog's models had a LiveBench coding result, and Anthropic Claude Fable 5 led with 98.3%, ahead of Anthropic Claude Fable 5.1 at 97.7%.
LiveBench is a benchmark that adds new questions on a schedule, so a model is less likely to have seen the answers in training. In its current release the coding category is bug-fix tasks taken from real merged pull requests on a frozen codebase, graded deterministically: a fix that breaks existing behaviour scores 0. We read the category score as a share of points.
Role in the index
LiveBench coding carries 15% of the Vibe Coding 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 Fable 5Anthropic | 98.3% | $8.9273 | 49,492 | 79.3 | $50.00 | |
| 2 | Claude Fable 5.1Anthropic | 97.7% | $5.2824 | 41,078 | 84.2 | $50.00 | |
| 3 | Claude Opus 5.5Anthropic | 96.3% | $5.0308 | 84,646 | 88.8 | $20.00 | |
| 4 | DeepSeek V4.1 Flash 1017DeepSeek | 96.0% | $0.1955 | 62,695 | 70.3 | $0.60 | |
| 5 | GPT-6.1 SolOpenAI | 96.0% | $0.4822 | 14,030 | 82.4 | $10.00 | |
| 6 | Claude Sonnet 5.5Anthropic | 94.7% | $4.0218 | 124,443 | 87.3 | $10.00 | |
| 7 | Grok 4.5xAI | 94.3% | $0.9592 | 15,315 | 58.6 | $6.00 | |
| 8 | GPT-6 AstraOpenAI | 93.7% | $3.3154 | 13,904 | 85.1 | $50.00 | |
| 9 | GPT-6 SolOpenAI | 93.1% | $0.9769 | 27,853 | 76.0 | $10.00 | |
| 10 | Claude Haiku 5.5Anthropic | 91.4% | $0.8738 | 138,052 | 70.7 | $0.50 | |
| 11 | Claude Opus 5Anthropic | 90.8% | $3.9218 | 35,875 | 83.0 | $25.00 | |
| 12 | DeepSeek V4.1 FlashDeepSeek | 90.2% | $0.2377 | 73,074 | 62.5 | $0.78 | |
| 13 | Muse Spark 1.3Meta | 90.2% | $1.3716 | 34,886 | 57.4 | $4.25 | |
| 14 | Gemini 4 ArgonGoogle | 89.7% | $5.0763 | 115,041 | 81.0 | $10.00 | |
| 15 | GPT-5.6 SolOpenAI | 89.7% | $3.7688 | 23,972 | 73.5 | $20.00 | |
| 16 | Grok 4.6xAI | 89.1% | $1.8584 | 25,609 | 65.7 | $6.00 | |
| 17 | Qwen3.8 FlashQwen | 87.4% | $0.1061 | 35,754 | Not benchmarked | $0.47 | |
| 18 | Qwen3.8 27BQwen | 86.8% | $0.69 | 43,464 | 74.2 | $2.55 | |
| 19 | GPT-5.6 TerraOpenAI | 86.2% | $2.7147 | 33,438 | 63.9 | $12.00 | |
| 20 | Step 5 PreviewStepFun | 85.6% | $1.0953 | 43,804 | 61.1 | $2.70 | |
| 21 | GPT-5.5OpenAI | 85.1% | $4.2076 | 18,415 | 71.3 | $30.00 | |
| 22 | GPT-6 LunaOpenAI | 84.5% | $0.0951 | 75,610 | 57.3 | $0.50 | |
| 23 | Grok 4.7xAI | 84.5% | $4.8799 | 59,183 | 66.7 | $6.00 | |
| 24 | GLM 5.3Z.ai | 82.8% | $2.1384 | 45,039 | 70.6 | $4.40 | |
| 25 | Gemini 3.8 FlashGoogle | 82.2% | $2.2964 | 111,470 | 61.6 | $3.75 | |
| 26 | MiMo-V2.6-ProXiaomi | 81.5% | $0.2239 | 82,405 | 65.4 | $0.87 | |
| 27 | GLM 5.3 FlashZ.ai | 80.6% | $0.1696 | 29,839 | 57.7 | $0.50 | |
| 28 | Gemini 3.7 FlashGoogle | 80.5% | $1.7317 | 58,180 | 58.0 | $3.75 | |
| 29 | Muse Spark 1.2Meta | 78.2% | $1.4657 | 35,890 | 55.7 | $4.25 | |
| 30 | Claude Opus 4.8Anthropic | 77.3% | $4.1517 | 55,126 | 65.0 | $25.00 | |
| 31 | Claude Sonnet 5Anthropic | 74.7% | $3.4831 | 77,338 | 57.7 | $10.00 | |
| 32 | Muse Spark OpenMeta | 73.6% | $1.2525 | 26,529 | 47.4 | $4.25 | |
| 33 | GPT-5.6 LunaOpenAI | 73.6% | $1.8232 | 35,428 | 58.0 | $1.20 | |
| 34 | MiniMax M3MiniMax | 70.9% | $0.4551 | 38,007 | 39.2 | $1.20 | |
| 35 | Claude Sonnet 4.6Anthropic | 66.6% | $2.6287 | 37,618 | 57.3 | $15.00 | |
| 36 | Nemotron 3.5 Super VLNVIDIA | 65.5% | $0.3307 | 46,228 | Not benchmarked | $0.50 | |
| 37 | T InklingThinking Machines | 64.0% | $1.8766 | 30,579 | 31.5 | $4.05 | |
| 38 | Qwen3.8 Max (0902)Qwen | 60.3% | $1.9869 | 167,999 | 62.4 | $6.00 | |
| 39 | Mistral Large 4Mistral AI | 60.0% | $0.9826 | 45,501 | 56.4 | $4.18 | |
| 40 | Kimi K3Moonshot AI | 59.3% | $2.0614 | 18,869 | 62.3 | $15.00 | |
| 41 | Gemma 4 31BGoogle | 49.0% | $0.2326 | 11,719 | 39.0 | $0.40 | |
| 42 | Claude Haiku 4.5Anthropic | 48.1% | $0.494 | 19,224 | 35.6 | $5.00 | |
| 43 | DeepSeek V4 Flash Vision ExpDeepSeek | 37.6% | $0.0401 | 18,084 | Not benchmarked | $1.32 |
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 21, 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 | LLM API Integration | Agent Loop | Billing & Limits | Cloud Infrastructure | Product & UX | Memory & Tools |
|---|---|---|---|---|---|---|---|
| Anthropic Claude Fable 5 | 98.3% | 100.0% | 100.0% | 100.0% | 100.0% | 83.3% | 100.0% |
| Anthropic Claude Fable 5.1 | 97.7% | 100.0% | 100.0% | 100.0% | 100.0% | 83.3% | 100.0% |
| Anthropic Claude Opus 5.5 | 96.3% | 97.5% | 96.0% | 100.0% | 100.0% | 88.9% | 97.5% |
| DeepSeek V4.1 Flash 1017 | 96.0% | 87.5% | 100.0% | 100.0% | 100.0% | 88.9% | 100.0% |
| OpenAI GPT-6.1 Sol | 96.0% | 83.3% | 90.0% | 100.0% | 100.0% | 100.0% | 100.0% |
| Anthropic Claude Sonnet 5.5 | 94.7% | 100.0% | 100.0% | 83.3% | 98.3% | 87.2% | 100.0% |
| SpaceXAI Grok 4.5 | 94.3% | 75.0% | 100.0% | 100.0% | 94.4% | 94.4% | 100.0% |
| OpenAI GPT-6 Astra | 93.7% | 75.0% | 90.0% | 100.0% | 100.0% | 94.5% | 100.0% |
| OpenAI GPT-6 Sol | 93.1% | 100.0% | 100.0% | 100.0% | 100.0% | 66.7% | 100.0% |
| Anthropic Claude Haiku 5.5 | 91.4% | 100.0% | 100.0% | 95.8% | 83.3% | 77.8% | 100.0% |
| Anthropic Claude Opus 5 | 90.8% | 100.0% | 100.0% | 100.0% | 100.0% | 66.7% | 75.0% |
| DeepSeek V4.1 Flash | 90.2% | 87.5% | 100.0% | 75.0% | 88.9% | 88.9% | 100.0% |
| Meta Muse Spark 1.3 | 90.2% | 75.0% | 100.0% | 95.8% | 100.0% | 72.2% | 100.0% |
| Google Gemini 4 Argon | 89.7% | 100.0% | 80.0% | 100.0% | 100.0% | 77.8% | 83.3% |
| OpenAI GPT-5.6 Sol | 89.7% | 75.0% | 100.0% | 100.0% | 100.0% | 66.7% | 100.0% |
| SpaceXAI Grok 4.6 | 89.1% | 75.0% | 80.0% | 95.8% | 100.0% | 83.3% | 100.0% |
| Qwen3.8 Flash | 87.4% | 100.0% | 80.0% | 95.8% | 91.7% | 66.7% | 100.0% |
| Qwen3.8 27B | 86.8% | 87.5% | 100.0% | 79.2% | 86.1% | 72.2% | 100.0% |
| OpenAI GPT-5.6 Terra | 86.2% | 87.5% | 100.0% | 79.2% | 100.0% | 72.2% | 75.0% |
| StepFun Step 5 Preview | 85.6% | 100.0% | 100.0% | 70.8% | 77.8% | 72.2% | 100.0% |
| OpenAI GPT-5.5 | 85.1% | 87.5% | 100.0% | 75.0% | 75.0% | 77.8% | 100.0% |
| OpenAI GPT-6 Luna | 84.5% | 75.0% | 90.0% | 100.0% | 100.0% | 83.3% | 50.0% |
| SpaceXAI Grok 4.7 | 84.5% | 37.5% | 100.0% | 100.0% | 100.0% | 66.7% | 100.0% |
| Z.ai GLM 5.3 | 82.8% | 75.0% | 100.0% | 75.0% | 83.3% | 66.7% | 50.0% |
| Google Gemini 3.8 Flash | 82.2% | 87.5% | 80.0% | 75.0% | 100.0% | 72.2% | 75.0% |
| Xiaomi MiMo-V2.6-Pro | 81.5% | 50.0% | 100.0% | 100.0% | 90.0% | 72.2% | 72.5% |
| Z.ai GLM 5.3 Flash | 80.6% | 50.0% | 80.0% | 100.0% | 83.3% | 50.0% | 75.0% |
| Google Gemini 3.7 Flash | 80.5% | 100.0% | 80.0% | 75.0% | 83.3% | 66.7% | 75.0% |
| Meta Muse Spark 1.2 | 78.2% | 62.5% | 60.0% | 75.0% | 91.7% | 77.8% | 100.0% |
| Anthropic Claude Opus 4.8 | 77.3% | 50.0% | 60.0% | 75.0% | 100.0% | 66.7% | 50.0% |
| Anthropic Claude Sonnet 5 | 74.7% | 50.0% | 80.0% | 79.2% | 91.7% | 66.7% | 75.0% |
| Meta Muse Spark Open | 73.6% | 50.0% | 78.0% | 50.0% | 91.7% | 72.2% | 90.2% |
| OpenAI GPT-5.6 Luna | 73.6% | 87.5% | 80.0% | 29.2% | 100.0% | 61.1% | 75.0% |
| MiniMax M3 | 70.9% | 50.0% | 50.0% | 50.0% | 88.9% | 83.3% | 92.9% |
| Anthropic Claude Sonnet 4.6 | 66.6% | 100.0% | 40.0% | 50.0% | 66.7% | 50.0% | 75.0% |
| NVIDIA Nemotron 3.5 Super VL | 65.5% | 75.0% | 70.0% | 50.0% | 75.0% | 33.3% | 100.0% |
| Thinking Machines Inkling | 64.0% | 75.0% | 60.0% | 75.0% | 66.7% | 47.2% | 67.9% |
| Qwen3.8 Max (0902) | 60.3% | 50.0% | 60.0% | 50.0% | 33.3% | 50.0% | 75.0% |
| Mistral Large 4 | 60.0% | 68.3% | 56.0% | 45.8% | 47.2% | 55.6% | 96.4% |
| MoonshotAI Kimi K3 | 59.3% | 50.0% | 40.0% | 75.0% | 50.0% | 50.0% | 25.0% |
| Google Gemma 4 31B | 49.0% | 54.2% | 50.0% | 25.0% | 63.9% | 33.3% | 67.9% |
| Anthropic Claude Haiku 4.5 | 48.1% | 62.5% | 50.0% | 50.0% | 50.0% | 16.7% | 74.0% |
| DeepSeek V4 Flash Vision Exp | 37.6% | 75.0% | 40.0% | 25.0% | 48.3% | 0.0% | 50.0% |
The parts as LiveBench publishes them, for the same run as the result.