Tencent releases Hunyuan Hy4 flash, citing lower costs on internal task tests

Originally reported November 16, 2026. Tencent Hunyuan has released Hy4 flash, a language model with 224 billion total parameters—including 64 billion Engram parameters—and 12 billion activated parameters. Tencent says it supports multimodal input and a context length of 1 million tokens.

Tencent lists pricing at ¥0.50 per million input tokens and ¥2 per million output tokens, with a ¥0.02 rate when the cache is hit. The company says architectural and inference optimizations reduce the model’s KV-cache size and enable decoding speeds above 150 tokens per second. These are Tencent’s reported figures, not independent measurements.

Tencent releases Hunyuan Hy4 flash, citing lower costs on internal task tests (1)

On WorkIndex, an internal evaluation built around WorkBuddy tasks such as document editing, presentations and visual understanding, Tencent says Hy4 flash performed slightly better overall than the larger Hy4 preview. It estimates per-task costs at one-thirteenth of Hy4 preview’s and one-twentieth of Hy4’s. Tencent also reports that a blind test with 85 internal WorkBuddy users, covering 259 tasks across 27 categories, found user satisfaction close to Claude Haiku 5.5. Those results should not be read as independent benchmark findings.

Tencent releases Hunyuan Hy4 flash, citing lower costs on internal task tests (2)

Hy4 flash is available through Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis, according to the report. OpenRouter integration and an open-source release are planned, not yet completed.

Source: infoq.cn (originally published 2026-11-16 14:23 UTC+8)