PARSA-Bench - ASR: leaderboard
Metric: Median word error rate times 100 on 500 Persian Common Voice and ParsVoice utterances (0 is a perfect transcript, no upper limit); zero-shot, temperature 0, refusals (no Persian script) excluded; proprietary models on 100 items; lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 40 | #237 |
| 2 | Gemma 3n E4B | 50 | #856 |
| 3 | Gemma 3n E2B | 64 | #1137 |
| 4 | Qwen2.5-Omni-7B | 100 | #601 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=parsa-bench-asr · How It Works · Data refreshed daily, snapshot 2026-10-11.