Tool-Genesis (Direct) - Unit Tests (Soft): leaderboard
Metric: Unit-test score (0-1, times 100) over the standard tests: each test averages a key-path F1 and an embedding similarity between the generated tool's output and the expected output, single-pass generation (Direct), Tool-Genesis's tool-creation tasks over 86 executable MCP servers (508 tools, 9,441 unit tests): from an abstract requirement the model writes the tool interface schema and an executable MCP server implementation; temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 28.1 | #131 |
| 2 | GPT-4.1 | 26.1 | #240 |
| 3 | DeepSeek V3.2 | 19.5 | #198 |
| 4 | Qwen 3 32B | 17.6 | #424 |
| 5 | Qwen 3 14B | 15.4 | #524 |
| 6 | Kimi K2 | 14.4 | #236 |
| 7 | Qwen 3 235B A22B 2507 Instruct | 14.3 | #291 |
| 8 | GPT-4.1 Mini | 12.7 | #346 |
| 9 | Qwen 3 30B A3B 2507 Instruct | 9.3 | #464 |
| 10 | GPT-4o | 8.9 | #333 |
| 11 | Gemini 3 Flash (Preview) | 8.4 | #78 |
| 12 | Claude 3.5 Haiku | 0.7 | #553 |
| 13 | Qwen 3 8B | 0.1 | #667 |
| 14 | Qwen 3 4B | 0 | #823 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=tool-genesis-direct-unit-tests-soft · How It Works · Data refreshed daily, snapshot 2026-10-11.