RobustBench-TC - Typo: leaderboard

RobustBench-TC (2026) tests tool-calling models under deployment noise grounded in real tool-calling failures, using 199 single-turn samples drawn from BFCL, API-Bank, RoTBench, ToolAlpaca and ToolEyes and scored by deterministic rule-based parsers. This board is tool-call accuracy in percent under keyboard typos in the user query; higher is better.

Metric: Tool-call accuracy (%) under the Typo perturbation (observation component: user query (character-level keyboard noise)); RobustBench-TC samples drawn from BFCL V3 single-turn, API-Bank, RoTBench, ToolAlpaca and ToolEyes, each scored by its source benchmark's strict scorer after a format-tolerant tool-call parse, temperature 0 (DeepSeek-R1-Distill at 0.6), Qwen3-family models with thinking disabled; higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1LoopTool-32B67.3
2Qwen 3 32BNon-reasoning65.3#972
3Qwen 3 14BNon-reasoning64.3#1269
4ToolRL-Qwen2.5-1.5B62.8#2299
5O4 Mini61.8#368
6Qwen 3.5 9BNon-reasoning61.3#828
7MUA-RL-8B60.3#1640
8MUA-RL-32B59.8
9Qwen 3 8BNon-reasoning58.8#1873
10LoopTool-8B58.3#1473
11TL-CodeLLaMA-257.8
12DeepSeek R1 Distill Qwen 14B56.8#2771
13ToolRL-Qwen2.5-3B56.3
14MUA-RL-14B56.3#1658
15Qwen 2.5 3B Instruct55.8#5392

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=robustbench-tc-typo · How It Works · Data refreshed daily, snapshot 2026-11-21.