Change language to
0:00

Tencent has launched Hy4 flash, a cheaper AI model for everyday office work such as research, document processing, spreadsheets, and slide decks.

The model’s pitch is cost per task. According to Superpower Daily, Hy4 flash costs about US$0.070 per million input tokens and US$0.281 per million output tokens, roughly 92% and 89% less than full Hy4. Firms in the UAE and Saudi Arabia scaling AI tools across staff will watch that number closely. All prices here are in US dollars; no AED or SAR price has been given.

Tencent announced the model on 16 November. Hy4 flash is aimed at most everyday tasks, including deep research, document processing, PowerPoint generation, simple front-end displays, and computer use. The full Hy4 remains the pick for complex work such as long-horizon software engineering.

Per million tokensHy4 flashHy4
InputUS$0.070US$0.90
OutputUS$0.281US$2.70
Cached inputUS$0.003US$0.045

Subscribe to our Newsletters for more Tech Stories

What’s inside

Hy4 flash has 224 billion total parameters, including 64 billion Engram parameters, but only 12 billion are active at a time. It accepts multimodal input and a context of up to 1M tokens. Decoding runs at more than 150 tokens per second, and its KV cache is 0.042 the size of Hy3’s and 0.167 of Hy4’s. Reinforcement learning also cuts average task output by 50%.

For developers, the API identifier is hy4-flash, according to Unite.AI. It has a 960,000-token maximum input and 64,000-token maximum output, and supports the OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages protocols.

Tencent’s numbers

On Tencent’s internal WorkIndex benchmark, Hy4 flash slightly outperforms Hy4 Preview (770 billion total, 49 billion active) overall. Tencent says its per-task cost is 0.077 of Hy4 Preview’s, 0.05 of Hy4’s, and 0.125 of DeepSeek-V4.1-Flash-1017’s at the Max tier with peak-hour pricing.

In a blind test, 85 Tencent WorkBuddy users rated outputs across 259 tasks in 27 categories. Hy4 flash reached satisfaction close to Claude Haiku 5.5, with the lowest per-task cost and shortest completion time of the models tested. Both results come from company-run tests; independent comparisons aren’t available yet.

Availability and background

Hy4 flash is available in Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima, and Marvis, with OpenRouter integration coming soon. Until 30 November, it’s free on WorkBuddy and 50% off standard pricing through TokenHub.

Hy4 itself was released on 4 November. Tencent says it ranks second among Chinese models on the Artificial Analysis Intelligence Index and first among open-source models on the Agent Arena leaderboard. Tencent also plans to open-source Hy4 flash, with no date given.

Tencent Cloud has been expanding capacity in Riyadh. The region also has its own model builders, such as the K2 family from MBZUAI.