Hy4

Hy4 is the official release of Tencent Hunyuan’s next-generation LLM (OpenRouter id tencent/hy4). Versus the earlier Hy4 Preview (MoE, ~770B total / 49B active, coding-agent oriented), the official card claims comprehensive gains in agentic ability and long-horizon productivity workloads, with ~1,048,576-token context.

OpenRouter listed the official model around 2026-11-04 12:20 Asia/Shanghai. Same-day discussion tied it to Wuzhen open-source messaging and strong preview leaderboard/cost signals. An earlier radar pass (04:32) correctly treated Hy4 as a teaser; this scan inserts on ship evidence.

Problem

Teams that want Tencent Hunyuan’s latest official tier on an open router—especially for agents/tools and very long context—can call tencent/hy4 instead of relying on Preview-only messaging.

Differentiation

• vs Hy4 Preview: official release stresses agentic/long-horizon gains; Preview remains listed separately.
• vs other open models: named Hunyuan product launch with Wuzhen open-source narrative.

Limits

• Exact open-weight hosting, license, and any Flash SKU timing need official Hunyuan/ModelScope sources.
• Pricing/rate limits follow OpenRouter/Tencent Cloud pages.
• Preview leaderboard ≠ final official benchmarks.

Who it is for

Builders of coding agents, tool orchestration, and long-document productivity apps who want Hunyuan’s latest official tier.

Getting started

1. Open the OpenRouter Hy4 page and call tencent/hy4 (compare with tencent/hy4-preview).
2. Watch Hunyuan / Wuzhen announcements for open weights and any Flash SKU.
3. A/B on your own long-horizon agent workloads before swapping defaults.