grok-build: 31.0 | Keygen Bench
Rank 74 of 77. Best of 3 attempts writing a keygen tune in FastTracker II. Listen to it in the tracker.
Keygen Bench asks AI models to compose a keygen-style chiptune in FastTracker II, with no network and a Bash tool, three independent attempts each at its highest declared reasoning tier. The score measures pitched clarity, development and dynamics from the XM module and a trusted FT2 render, with signal integrity and loop continuity checks. Models are ranked by their best of three attempts.
Ranking, best of 3 attempts| Rank | Model | Maker | Best score | Attempt scores |
|---|
| 1 | grok-4-6 | xAI | 76.3 | 70.9 / 71.6 / 76.3 |
| 2 | glm-5-3 | Zhipu | 75.5 | 65.5 / 75.5 / 69.3 |
| 3 | claude-sonnet-5-5 | Anthropic | 75.4 | 75.4 / 60.3 / 67.7 |
| 4 | gpt-6-sol | OpenAI | 75.4 | 75.4 / 70.1 / 67.1 |
| 5 | mimo-v2.6-pro | Xiaomi | 75.3 | 75.3 / 58.1 / 70.2 |
| 6 | gpt-5.6-sol | OpenAI | 74.6 | 74.6 / 70.2 / 73.3 |
| 7 | glm-5.2 | Zhipu | 73.5 | 70.6 / 67.0 / 73.5 |
| 8 | claude-fable-5-1 | Anthropic | 73.0 | 72.2 / 72.3 / 73.0 |
| 9 | gpt-6.1-sol | OpenAI | 72.7 | 72.2 / 71.9 / 72.7 |
| 10 | claude-fable-5 | Anthropic | 72.4 | 54.0 / 72.4 / 54.4 |
| 11 | grok-4-7 | xAI | 72.3 | 72.3 / 61.2 / 65.6 |
| 12 | claude-opus-4-7 | Anthropic | 72.1 | 33.5 / 69.6 / 72.1 |
| 13 | claude-opus-5 | Anthropic | 72.1 | 72.1 / 71.5 / 67.7 |
| 14 | gemini-3.5-flash | Google | 72.1 | 72.1 / 63.4 / 40.1 |
| 15 | deepseek-v4-1-flash | DeepSeek | 72.0 | 69.4 / 72.0 / 67.3 |
| 16 | claude-opus-5-5 | Anthropic | 71.6 | 63.0 / 71.6 / 64.5 |
| 17 | gpt-6-astra | OpenAI | 71.6 | 71.6 / 68.7 / 71.5 |
| 18 | claude-opus-4-8 | Anthropic | 70.8 | 44.9 / 70.8 / 69.2 |
| 19 | gpt-5.6-terra | OpenAI | 70.6 | 56.7 / 70.6 / 50.9 |
| 20 | claude-opus-4-6 | Anthropic | 70.5 | 33.8 / 48.9 / 70.5 |
| 21 | gpt-5.5 | OpenAI | 70.5 | 42.7 / 70.5 / 21.2 |
| 22 | gpt-6-luna | OpenAI | 70.5 | 47.5 / 35.3 / 70.5 |
| 23 | step-3.7-flash | StepFun | 70.5 | 70.5 / 57.7 / 25.2 |
| 24 | kimi-k3 | Moonshot AI | 70.3 | 70.3 / 47.4 / 65.6 |
| 25 | claude-haiku-5-5 | Anthropic | 70.2 | 66.7 / 70.2 / 60.1 |
| 26 | deepseek-v4-pro | DeepSeek | 69.8 | 69.8 / 66.7 / 41.4 |
| 27 | muse-spark-1.3 | Meta | 69.8 | 69.8 / 54.8 / 69.2 |
| 28 | claude-sonnet-5 | Anthropic | 69.6 | 69.6 / 54.0 / 66.4 |
| 29 | mistral-large-4 | Mistral AI | 69.6 | 39.2 / 69.6 / 42.3 |
| 30 | gemini-3.7-flash | Google | 69.4 | 69.4 / 56.0 / 26.8 |
| 31 | grok-4-5 | xAI | 69.2 | 33.7 / 29.5 / 69.2 |
| 32 | hy4-preview | Tencent | 68.8 | 65.2 / 68.8 |
| 33 | claude-haiku-4-5-20251001 | Anthropic | 67.8 | 67.8 / 3.4 / 11.4 |
| 34 | deepseek-v4-flash | DeepSeek | 67.7 | 67.7 / 51.1 |
| 35 | grok-composer-2.5-fast | xAI | 67.6 | 50.9 / 26.6 / 67.6 |
| 36 | kimi-k2.7-code | Moonshot AI | 67.2 | 33.4 / 64.8 / 67.2 |
| 37 | swe-2 | Cognition | 67.1 | 66.2 / 57.8 / 67.1 |
| 38 | hy3 | Tencent | 66.9 | 28.6 / 32.5 / 66.9 |
| 39 | gemini-3-flash-preview | Google | 66.7 | 61.7 / 29.9 / 66.7 |
| 40 | gpt-5-4 | OpenAI | 66.2 | 36.8 / 66.2 / 47.9 |
| 41 | gemini-3.8-flash | Google | 66.1 | 37.9 / 52.2 / 66.1 |
| 42 | swe-1-7-lightning | Cognition | 65.9 | 65.9 / 22.1 / 59.0 |
| 43 | gemini-3.6-flash | Google | 65.3 | 31.3 / 65.3 / 49.6 |
| 44 | gpt-5.6-luna | OpenAI | 64.8 | 64.8 / 27.7 / 24.4 |
| 45 | gpt-5-3-codex | OpenAI | 64.0 | 64.0 / 28.5 / 52.5 |
| 46 | minimax-m3 | MiniMax | 62.4 | 62.4 / 28.7 / 16.2 |
| 47 | muse-glimmer-30b | Meta | 61.9 | 61.9 / 52.1 / 50.7 |
| 48 | grok-4.20-0309-reasoning | xAI | 61.7 | 57.3 / 38.8 / 61.7 |
| 49 | claude-sonnet-4-6 | Anthropic | 61.5 | 61.5 / 37.1 / 32.4 |
| 50 | glm-5-3-flash | Zhipu | 61.3 | 34.4 / 61.3 |
| 51 | qwen3.5-397b-a17b | Alibaba | 58.8 | 58.8 / 15.5 / 31.5 |
| 52 | kimi-k2-6 | Moonshot AI | 57.5 | 34.2 / 57.5 / 32.4 |
| 53 | hermes-4-405b | Nous Research | 57.0 | 57.0 / 40.4 |
| 54 | nemotron-3-ultra | NVIDIA | 55.3 | 52.6 / 55.3 / 24.7 |
| 55 | command-a-plus-05-2026 | Cohere | 55.1 | 33.0 / 55.1 |
| 56 | north-mini-code | Cohere | 54.2 | 54.2 / 9.1 |
| 57 | command-a-reasoning-08-2025 | Cohere | 53.8 | 53.8 |
| 58 | mercury-2.5 | Inception | 52.4 | 52.4 / 40.3 |
| 59 | claude-opus-4-5-20251101 | Anthropic | 43.6 | 35.7 / 19.3 / 43.6 |
| 60 | minimax-m2.7 | MiniMax | 43.6 | 35.0 / 43.6 / 43.4 |
| 61 | grok-4.20-0309-non-reasoning | xAI | 41.6 | 9.7 / 41.6 |
| 62 | qwen3.7-plus | Alibaba | 41.1 | 6.1 / 22.7 / 41.1 |
| 63 | gpt-5-4-mini | OpenAI | 40.9 | 28.4 / 40.9 / 23.2 |
| 64 | grok-4.3 | xAI | 39.6 | 39.6 / 33.9 / 17.5 |
| 65 | inkling | Thinking Machines Lab | 38.1 | 29.9 / 38.1 / 26.8 |
| 66 | grok-build-0.1 | xAI | 36.6 | 31.1 / 36.6 / 28.8 |
| 67 | gemini-3.1-pro-preview | Google | 36.2 | 30.8 / 21.6 / 36.2 |
| 68 | kimi-k2-7 | Moonshot AI | 34.4 | 34.4 / 25.5 / 24.0 |
| 69 | qwen3.8-flash | Alibaba | 34.4 | 34.4 / 20.6 / 32.1 |
| 70 | gpt-oss-120b | OpenAI | 33.9 | 29.1 / 33.9 |
| 71 | gpt-oss-20b | OpenAI | 33.8 | 33.8 / 17.3 |
| 72 | qwen3.8-27b | Alibaba | 33.2 | 33.2 |
| 73 | nemotron-3.5-lightning | NVIDIA | 31.2 | 31.2 |
| 74 | grok-build | xAI | 31.0 | 31.0 / 23.2 / 20.9 |
| 75 | nemotron-3-super-120b-a12b | NVIDIA | 30.0 | 12.0 / 13.5 / 30.0 |
| 76 | swe-1-6 | Cognition | 28.3 | 1.8 / 28.3 / 15.3 |
| 77 | muse-spark-1.2 | Meta | 25.3 | 21.6 / 25.3 / 19.5 |
Rankings | Tracker | Scoring | Support