Coverage confirms Hy4's open-source release with 770B parameters and 49B active per token
ConcludedRan 1 dayin the edition since 31 Aug 20266 sourcesupdated 31 Aug 2026
Tencent Hy4 Preview open-weight model release
The gist
Latest: Tencent Hy4 Preview: 770B MoE Open-Weight Model Benchmarked
Hy4 is among the largest open-weight models published, making a 770B-parameter MoE model available for download. Its coding benchmark results are competitive with several leading closed models, while graduate-level reasoning lags behind the top frontier tier.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Tencent released Hy4 Preview on August 29, 2026, as an open-weight mixture-of-experts model with 770B total parameters and 49B active per token, supporting a 1M-token context window, up from Hy3's 256K. Weights are available on GitHub, Hugging Face, ModelScope, and GitCode/AtomGit. Benchmark analyses show large coding gains over Hy3: the DeepSWE score rose from 28.00 to 64.30 and SWE-bench Multilingual from 75.80 to 82.90. At 'High' thinking level, Hy4 scores 92.30 on GPQA Diamond and 55.40 on HLE; in text-only, no-tools mode, HLE drops to 43.4, behind Claude Opus 5 (54.9) and GPT 5.6 Sol (49.5). On enterprise benchmarks, Hy4 scores 78.6 on BankerToolBench, ahead of DeepSeek V4 Pro and Kimi K3 but behind GPT 5.6 Sol and Claude Opus 5. Tencent's deployment FAQ states that open-source release does not mean unrestricted commercial use and that preview status is not production approval.
How it developed
First in the edition · 31 August 2026
Tencent released Hy4 Preview on August 29, 2026, a 770B mixture-of-experts open-weight model with 49B active parameters per token and a 1M-token context window, with weights published on GitHub and Hugging Face.
Benchmark analyses from DataLearnerAI show large coding gains over its predecessor Hy3, with DeepSWE rising from 28.00 to 64.30 and SWE-bench Multilingual from 75.80 to 82.90. In text-only, no-tools mode, HLE reaches 43.4, behind Claude Opus 5 at 54.9 and GPT 5.6 Sol at 49.5. Tencent's deployment FAQ states that open-source release does not mean unrestricted commercial use and that preview status is not production approval.
Sources
Simon Willison
1 more source
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.