Skip to content
Brief The AI-First Web ·

Tencent launches Hy4 flash, a cheaper AI model for office tasks

Tencent says the new model costs one-twentieth as much per task as the full Hy4 in its own tests.

In brief
  • Tencent launched Hy4 flash on Monday. It says a task costs one-twentieth as much as on the full Hy4 model.
  • The cost and quality results come from Tencent's own tests. The company plans to release the model's weights but gave no date.

TechNode reported that Tencent launched Hy4 flash on Monday. The model is meant for everyday work such as research, making slides and editing spreadsheets. Tencent said that on its own WorkIndex benchmark, a task cost one-twentieth as much as on the full Hy4 model. List prices for its API (the paid access developers call from code) are $0.07 per million input tokens and $0.281 per million output tokens. Cached input costs $0.003 per million tokens.

Tencent said the model takes more than text as input and can handle up to one million tokens at once. It has 224 billion parameters and uses 12 billion for each step. In a blind test, 85 WorkBuddy users rated outputs across 259 tasks. Tencent said satisfaction was close to Claude Haiku 5.5, with the lowest cost and shortest completion time of the models tested. The report does not name the other models. It gives no release date for the weights, and no end date for the half-price API offer. WorkBuddy users get free access through November 30.

For teams that pick models for routine office work, this adds a low-priced option with published rates. The cost and quality claims come from Tencent's own tests, so buyers will want to run their own checks. If Tencent releases the weights as planned, teams could also run the model themselves.

A WebPulse Brief: a short report of an important event, written by the WebPulse Newsroom with AI assistance and checked against the reporting below. When there is more to explain, we follow up with a full story. How we use AI.

Reporting: TechNode, Tech in Asia.