Tool-Genesis (Code-Agent) - Schema F1: leaderboard
Metric: Schema-F1 (0-1, times 100): F1 of the predicted tool interfaces against the reference after maximum-weight matching on embedding similarity of tool name and argument schema, a ReAct-style agent loop of up to 10 think-act-observe steps with sandboxed execution to run and repair the artifacts (Code-Agent), 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: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 91.2 | #78 |
| 2 | Kimi K2 | 89.8 | #236 |
| 3 | GPT-5.1 | 86.7 | #131 |
| 4 | Claude 3.5 Haiku | 82.1 | #553 |
| 5 | Qwen 3 32B | 80.1 | #424 |
| 6 | Qwen 3 235B A22B 2507 Instruct | 72.2 | #291 |
| 7 | Qwen 3 14B | 70.7 | #524 |
| 8 | DeepSeek V3.2 | 70.2 | #198 |
| 9 | Qwen 3 30B A3B 2507 Instruct | 69.4 | #464 |
| 10 | GPT-4.1 | 69.1 | #240 |
| 11 | Qwen 3 8B | 64.6 | #667 |
| 12 | GPT-4.1 Mini | 64 | #346 |
| 13 | GPT-4o | 45.6 | #333 |
| 14 | Qwen 3 4B | 28.5 | #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-code-agent-schema-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.