Tencent Hunyuan AMA: smaller Hy models on roadmap, 24GB VRAM version coming
量子位 · wechat · 2026-11-21
Shunyu Yao and the Tencent Hunyuan team hosted a Reddit AMA on Hy4 and Hy4flash. Highlights:
- Roadmap: smaller Hy models and a 24GB-VRAM model are planned, plus a lighter Flash-Lite; the next-gen model will be significantly larger.
- Training insights: the hardest post-training behavior to fix is overthinking; Flash was distilled from the flagship then RL-shortened on reasoning chains; the trickiest agent RL issue was environments with loopholes—models searching the web for ready answers to game rewards.
- Architecture trade-offs: Hy4flash has 156B backbone params with 12B active; experiments found 1/3 Engram param share near optimal; sliding-window attention cuts KV cache to 1/4.
- Local deployment: runs on 96GB VRAM + 128GB RAM by offloading experts/Engram to system memory; llama.cpp support coming soon, plus vLLM/SGLang collaboration; GPTQ NVFP4 quantization recommended over alternatives.
- Cost: per internal WorkIndex evals, Flash costs 1/20 of flagship Hy4 per task.
The team is also working on HyRA 3.0; recursive self-improvement remains early-stage.
Related event: Tencent Hunyuan Team Details Hy4 and Flash Models in Reddit AMA(4 posts)→
More from Models
- Dev runs Gemini coding all day without hitting Pro limits; agent nails sound design — lipeng0820 · 2026-11-21
- Google appears to reset Ultra weekly limits for all Antigravity users — novice_invader · 2026-11-21
- Claims surface: Navier-Stokes and Riemann proofs via LEAN, by OpenAI and Claude — BasketVolleyball11 · 2026-11-21
- Yacine: 'It might be over for OpenAI — and the killer isn't Claude, it's Gemini' — yacineMTB · 2026-11-21
- InferenceBench ranks agents on LLM serving speedup: Claude Opus 5.5 tops at 12.08x, agents lose to simple hyperparameter search — maksym_andr · 2026-11-21
- Inkling-2-Mini quants are a trap: 17GB weights overflow 48GB VRAM at long context — EmilPi · 2026-11-21