ToolsAugust 28, 2026via Vercel Blog
Hy4 Preview now available on AI Gateway
Why it matters
A capable open-weight model gains distribution through a developer-friendly platform. Practitioners building multi-agent stacks get another routing option; the 1M context window and MoE efficiency matter for long-horizon tasks.
Key signals
- Hy4 Preview: 770B total parameters, 49B active per token (MoE)
- 1M token context window
- Available via Vercel AI Gateway with model routing, cost tracking, zero data retention option
- Supports coding agents: Claude Code, Cursor, OpenCode, Pi, Codex
- Provider pricing with no platform fee or markup
- Use case focus: long-horizon coding, document analysis, game dev, scientific reasoning
- Hy4 Preview: 770B total parameters, 49B active per token (MoE architecture)
- Open-source model from Tencent
- Available via Vercel AI Gateway with unified API, usage tracking, routing, failover
- No platform fee on inference, reflects provider pricing
- Integrates with coding agents: Claude Code, Codex, OpenCode, Cursor, Pi
- SDK integration: set model to 'tencent/hy4-preview'
The hook
Tencent's Hy4 (770B MoE) joins Vercel's AI Gateway—practitioners can now route coding agents through a unified API with cost tracking and failover.
is now available on AI Gateway.Hy4 Preview from Tencent
Hy4 Preview is an open-source Mixture-of-Experts model with 770B total parameters and 49B active per token, aimed at long-horizon coding, document analysis, game development, and scientific reasoning. It serves a context window of 1M tokens.
T…