Tencent Releases Hy4 Preview: 770B MoE with 1M Context
ChrisGPT · x · 2026-08-28
Tencent's Hunyuan team has released the Hy4 preview, a massive MoE model with 770B total parameters (49B active per token), a 1M context window, and Apache 2.0 licensing.
Key Benchmarks:
- Coding/Agent: Scores 85.4 on Terminal Bench 2.1, placing it in the frontier cluster. DeepSWE performance jumps significantly from Hy3 (28.0) to 64.3, though it still trails Kimi K3 (74.0) and Claude Opus 5 (74.7).
- Other Tests: 17.5 on ProgramBench (vs. Claude's 39.5), 53.3 on SWE Atlas Refactoring (vs. 60.0), and 43.4 on Humanity's Last Exam (vs. 53.2).
Internal Eval: Tencent's transparent benchmark sheet highlights areas where the model still lags. Notably, an internal blind test of 203 engineering tasks reported on GitHub showed Hy4 slightly outperforming GLM 5.3 and Kimi K3.
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
- NVIDIA and PrimeIntellect post-train Nemotron 3.5 Super VL, boosting AutomationBench-AA by ~70% — willcb · 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