Tencent open-sources Hy4 preview: 770B MoE with 1M context that helped train itself
aigclink · x · 2026-08-28
Tencent has open-sourced its new flagship Hy4 preview: a 770B-parameter MoE with 49B active parameters and 1M context, optimized for real work scenarios — coding, games, finance and scientific simulation.
The standout feature is a recursive self-improvement loop: for the first time, Hy4 preview participated in its own development, proposing improvements, running experiments, and adjusting training methods, data strategy and kernel optimization based on results.
The team stresses "upgrades, not price hikes" — the model is free in WorkBuddy for two weeks, with weights, blog and GitHub repo available.
Related event: Tencent Open-Sources Hunyuan Hy4 Preview: 770B MoE with 1M Context(28 posts)→
More from Models
- Nathan Lambert slams Meta's Muse Open License as 'trash' — natolambert · 2026-10-16
- Google launches Gemini 4 Argon to Ultra and Pro subscribers, touting physics and 3D skills — ShikharMurty · 2026-10-16
- Google gives X users early access to test Argon, not just companies — lyraxana · 2026-10-16
- Grok Voice 3.0 Tops Design Arena Speech Leaderboard, Up 7.4 Points — XFreeze · 2026-10-16
- Robotics LLMs tested on 16 tasks: GPT-6 Astra hits 17.2% success vs Fable's 9.1% — pabbeel · 2026-10-16
- Meta open-sources Muse Spark: top US open model, scores just below DeepSeek — bindureddy · 2026-10-16