Skip to content
Hot eventDeveloping

Tencent releases Hy4 Flash, a 295B-parameter MoE model

1 reports1 reporting sourcesUpdated 1 h ago

Start here

AI digest

Tencent has released Hy4 Flash, a 295B-parameter Mixture-of-Experts model with 21B active parameters, 192 experts and top-8 routing, aimed at reasoning, agentic workflows and production use. The model offers a default no-think mode plus low and high chain-of-thought modes, a 256K context window, and tool-calling that Tencent says generalizes across agent scaffoldings. Tencent also emphasizes grounded, anti-hallucination behavior, though that is the company's own characterization rather than a verified result.

Generated from reports · updated 55 min ago

Report timeline

Follow the reports to see the event from different sides.

November 16
  1. Hacker News · AI
    Tencent Hy4 Flash: 295B-parameter MoE model with 256K context and configurable reasoning

    Tencent's Hy4 Flash is a 295B-parameter Mixture-of-Experts model with 21B active parameters, 192 experts and top-8 routing, built for reasoning, agentic workflows and production use. It offers a default no-think mode plus low and high chain-of-thought modes, a 256K context window, and stable tool-calling that generalizes across agent scaffoldings, with Tencent emphasizing grounded, anti-hallucination behavior.

Heat for this event

Not enough continuous observations yet to draw a trend.