Running on Zero Agents Qwen3 FunASR Thai Benchmark 🎙 Run a Thai speech‑to‑text benchmark comparing two ASR models