
Tencent launched Hy4 Flash on November 16. The new model costs about seven cents per million input tokens. It can work with more than one type of input and has a context window of up to one million tokens. Tencent built it for tasks such as edits to decks, docs and sheets.
Hy4 Flash came out less than two weeks after the full Hy4 model. The first release notice was in Chinese. Tencent put out an English version on November 17. The model is live in the WorkBuddy app and on Tencent Cloud. Read Tencent’s first launch notice.
What Tencent Hy4 Flash costs
The standard rates are about $0.070 per million input tokens and $0.281 per million output tokens. Cached input costs $0.003 per million tokens. That third rate is for text the service can reuse. A team that sends the same set of rules with each prompt may pay less if the service can use the copy it has saved.
At the standard rates, sending 10 million new input tokens and getting back one million would cost about 98 cents. That sum leaves out any discounts. The size of the context window sets how much the model can see in a run. How much you pay will depend on the tokens you send and get back.
How Tencent tested the model
Flash has 224 billion parameters in all and uses 12 billion at a time. Tencent says its setup can write more than 150 tokens per second. The result is for the setup used in its own tests. Tencent says it cut the size of the cache the model uses as it writes. The Flash cache is one-sixth the size of Hy4’s. Tencent gives the model specs and rates in its English release.
Tencent built a test called WorkIndex from tasks found in its WorkBuddy app. The tasks include making slides, fixing sheets and working with images. Tencent says Flash cost one-twentieth as much per task as Hy4 in that test. It still points users to Hy4 for hard coding jobs with lots of steps.
The test also compares costs with a DeepSeek Flash model at a specific tier and peak-hour price. That price can differ from the rate a team pays for its own use. AI World Today’s guide to DeepSeek models explains where its Flash and Pro lines fit.
Where to use Hy4 Flash
Tencent says Flash is live in TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis. It is free to use in WorkBuddy through November 30. On TokenHub, the firm is charging half of its standard rates through the same date. It says OpenRouter access will come later.
Tencent also says it plans to release the model as open source. For now, the release gives users access to a service that Tencent runs. A release of the weights could give teams the choice to run the model on their own systems. The terms and the hardware it would need are still to come. Our DeepSeek guide explains how hosted access and downloads work for that model family.
Read also
Popular news

OpenAI Launches GPT-6.2 Astra and Sol for ChatGPT Work, Codex and the API

We asked four AIs to predict the GTA 6 story. They all think Lucia and Jason survive

Nokia and Partners Plan a Four-Country AI Gigafactory Bid

How to Calculate AI ROI: Formula, Example and AI Prompt

AI Acceptable-Use Policy Template for Small Businesses

