Audio2Tool - Direct Commands: leaderboard

Speech-driven tool calling on Audio2Tool (2026), smart car, smart home and wearable requests spoken by diverse synthetic voices with background noise and given to an end-to-end speech language model, on the direct commands tier (short direct commands with a single tool and no arguments). Exact match of the tool calls and all arguments, in percent; higher is better.

Metric: Exact match (%): share of queries whose predicted tool or ordered tool sequence and all arguments exactly match the reference after deterministic normalization, on the 1,080 Tier 1 direct commands (2 to 6 words, a single tool with no arguments) of Audio2Tool, with the spoken query (voice-cloned TTS speech mixed with background noise) given to an end-to-end speech language model, zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 Omni 30B A3B92.4#847
2Qwen2.5-Omni-7B79.5#1884
3Step-Audio-2-mini77.9
4Kimi-Audio75.4#1735
5Audio Flamingo 364.5

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

Interactive version: theaggregate.ai/benchmark?slug=audio2tool-direct-commands · How It Works · Data refreshed daily, snapshot 2026-11-21.