Here's what I found about **Tencent: Hy4 preview** on OpenRouter: **What it is:** A mixture-of-experts (MoE) model from Tencent with **49B active parameters out of 770B total**. It's designed for coding agents, complex tool-use workflows, and productivity tasks that require planning, context continuity, and sustained multi-step execution. **Key specs:** - **Context window:** 1M tokens - **Released:** Aug 28, 2026 - **Pricing:** $0.834 / $2.501 per 1M tokens (input/output) - **Provider:** Hosted by a single provider (Tencent Cloud) — OpenRouter forwards every request directly to it **Performance (Tencent Cloud provider):** - Input: $0.834/M, Output: $2.501/M, Cache read: $0.042/M - Latency: 3.35s (P50) - Throughput: 39 tps - Uptime: 100.00% **Availability (last 3 days):** - Uptime (3d): 99.99% - Availability (3d): 98.47% - OpenRouter Availability: 98.79% **Notable usage:** The top public app sending traffic to this model is **Command Code** (1.98B tokens), followed by **Claude Code** (Anthropic's agentic coding tool) — consistent with its positioning for coding agents and tool-use workflows. Source: [OpenRouter — Tencent: Hy4 preview](https://openrouter.ai/tencent/hy4-preview)