Fish Audio Details Inference Stack: 0.17 RTF on a Single H200 GPU
rohanpaul_ai · x · 2026-07-30
Fish Audio revealed the underlying logic of how it can afford to offer its voice model for free: extreme inference optimization.
Hardware & Stack: Every request runs on a single NVIDIA H200 GPU using their custom software stack (fish-scales-ops for fast FP8 math and a 'pingpong' scheduler to prevent idle time).
Metrics: This setup achieves a Real-Time Factor (RTF) of 0.17, meaning it generates audio about 6x faster than real time, outputting 125 audio tokens per second with first sound in 70ms. This high single-GPU throughput drastically reduces the cost per request, making the free tier commercially viable.
Related event: Fish Audio Launches Free TTS API Optimized for Single H200 GPU(2 posts)→
More from Infra
- NVIDIA launches RTX Spark for local AI on Windows PCs — ostrisai · 2026-10-16
- SkyRL ships Day-0 RL support for Beam with +62% LoRA decode throughput in vLLM — vllm_project · 2026-10-16
- DFlash2 delivers up to 3.03x decode-only speedup for Beam across five benchmarks — vllm_project · 2026-10-16
- vLLM's Beam serving kernels cut GB200 forward time from 8.15 to 6.47 ms at TP8 batch 128 — vllm_project · 2026-10-16
- Reflection's 501B MoE Beam hits 463 tok/s on 8x GB200 with NVFP4 at day 0 — zhyncs42 · 2026-10-16
- Gemma 4 31B QAT on one 24GB card: 46 tok/s plain, ~80 with draft model on RTX 4090 — Subject_Poet_3861 · 2026-10-16