Tencent releases Hy4 Preview, a 770B-parameter open-weight LLM with 1M token context
Hy4 is Tencent’s next-generation text-only model, offering a fourfold jump in context length and a 2.6x increase in total parameters over its predecessor, Hy3.
1 source · cross-referenced
- Tencent released Hy4 Preview, an open-weight text-only LLM with 770B total parameters and 49B active parameters.
- Hy4 increases context length to 1M tokens, up from Hy3’s 256K tokens.
- The model is available on Hugging Face with a 1.56TB download size.
- Hy4 supports two reasoning effort modes: default "high" and optional "no_think".
Tencent has released Hy4 Preview, an open-weight text-only large language model with 770B total parameters and 49B active parameters. The model is positioned as a successor to Hy3, which launched in July with 295B total parameters and 21B active parameters.
Hy4 expands the context window to 1M tokens, a fourfold increase over Hy3’s 256,000-token context. The model is hosted on Hugging Face with a download size of 1.56TB.
The model’s chat template reveals two supported reasoning effort levels: "high" (the default) and "no_think" (a mode that disables internal reasoning traces).
Early testing by Simon Willison using the prompt "Generate an SVG of a pelican riding a bicycle" produced a detailed reasoning trace that iterated on visual details such as adding a helmet or sunglasses, demonstrating the model’s internal deliberation process.
- Oct 13, 2026 · Google DeepMind — Blog
Google DeepMind releases Gemini 4 Argon with domain-leading performance and multi-domain benchmarks
Trust79 - Oct 9, 2026 · Hugging Face
Hugging Face publishes Falcon ASR, a 1.6B-parameter Arabic speech recognition model with Emirati focus
Trust84 - Oct 8, 2026 · OpenAI — News
OpenAI disrupts two AI-enabled influence operations using its models
Trust79