Self-hosted LLM with SLERP-merged GRPO experts outperforms larger baseline, serving half of production traffic
t-tech · hf · 2026-09-02
t-tech shares how they trained separate GRPO experts on their real production request mix, then merged them via SLERP into a smaller self-hosted LLM.
The merged model outperforms a much larger baseline on instruction following, function calling, and internal tasks, while serving half of platform traffic at lower cost—demonstrating a practical path of post-training to your own traffic distribution.
More from Infra
- SGLang Summit sponsors grow from 11 to 36 in one week; Lip-Bu Tan, Lilian Weng to speak — BanghuaZ · 2026-10-18
- Yacine jokes his datacenter KPI is gigawatt-hours: get paged if not burning enough power — yacineMTB · 2026-10-18
- 1.68x inference speedup drops, directly accelerating GRPO post-training — yacineMTB · 2026-10-18
- OpenRouter is only a fraction of the total addressable market, poster argues — sloppenheimer · 2026-10-18
- Developer builds change alerts for LLM prices and retirements across Bedrock, Azure and Vertex — jsandhu0780 · 2026-10-18
- 12.3M tokens in 27 minutes for $0.14: agent session with 99.6% cache hit — teortaxesTex · 2026-10-18