Tencent open-sources HY4: 770B params, 49B active, 1M context, video-capable
heyshrutimishra · x · 2026-11-20
Tencent has released the weights of its new flagship HY4 model on Hugging Face, free to download: 770B total params with 49B active per token, 1M-token context, and a single model that reads text, images and video. Instruct, fp8 and base versions are all out.
The eye-catching model card claim: paired with Tencent's own agent harness, HY4 optimized the pre-training setup for smaller models and autonomously trained a variant that beat their official baseline on multiple benchmarks — the model helped build the next models. The card is candid about remaining gaps to top closed models in research-level math, long-horizon coding and self-improvement.
After DeepSeek V4, Kimi K3 and GLM 5.3, this is another Chinese lab shipping frontier-scale open weights, narrowing the open/closed gap.
More from coding & agent
- llms.txt Protocol Goes Mainstream as Google Antigravity Adopts It Two Years On — jeremyphoward · 2026-11-22
- He built a feed only agents can read: no UI, semantic matching, agent-rated posts — imdogefather · 2026-11-22
- Claude Code Debuts Experimental Code Mode as AI-Driven Tooling Gains Momentum — lxfater · 2026-11-22
- ClaudeAI Weekly: Single-HTML-File Arms Race, 8.9B-Token Strawberry, Post-IPO Anthropic — ClaudeAI-mod-bot · 2026-11-22
- Agno's AgentOS turns any agent into a production FastAPI app with MCP, RBAC and tracing — pritisinghhhh · 2026-11-22
- Chinese devs vent: app store rejections and copycats are just the norm — tinyfool · 2026-11-22