Tencent released Hy4 Preview, an open-weight text-only LLM with 770B total parameters, 49B active parameters, and a 1M token context window, weighing 1.56TB on Hugging Face. The model is substantially larger than Tencent's Hy3 from July, which had 295B parameters, 21B active, and 256,000 context. Its chat template supports two reasoning effort levels, "high" (default) and "no_think". Testing via OpenRouter showed a reasoning trace using truncated English.
No score is assigned. Sources and their independence are shown in the citation chain below.