Hy4 by Tencent added to BenchGecko
Tencent's flagship Hy4 adds image input and computer use · $0.834 input and $2.501 output per million tokens at list price · 1M token context.
What was added
Hy4 by Tencent is now listed on BenchGecko. Tencent released it on November 4, 2026 as the full version of Hy4 preview, which BenchGecko has tracked since August. It is the first Hy model to accept image input, and Tencent says it can also operate command line and graphical interfaces, including specialist software such as Blender and CAD tools.
Context and output
The API accepts a context window of 1,048,576 tokens, the same as Hy4 preview. The Tencent endpoint on OpenRouter allows up to 65,536 output tokens per response, up from 64,000 on the preview. Input is text and images, and output is text.
Pricing
Tencent lists Hy4 at $0.834 per million input tokens, $0.042 per million cached input tokens and $2.501 per million output tokens, the same price as Hy4 preview. At launch the Tencent endpoint on OpenRouter carries a 50% discount, so it currently charges $0.417 input, $0.021 cached input and $1.2505 output per million tokens. A task with 1 million input tokens and 500,000 output tokens costs about $2.08 at list price and about $1.04 at the launch rate, before any cache discount.
Availability
Hy4 is available through Tencent Cloud and OpenRouter, and inside Tencent products including Yuanbao, CodeBuddy and WorkBuddy. Tencent says it will publish the weights together with a system card, so BenchGecko lists Hy4 as a proprietary API model until they are released. Tencent also plans a smaller Hy4 flash model focused on cost and speed.
Benchmarks
Tencent calls Hy4 one of the strongest open weight models available, while stating that a notable gap remains to top proprietary models on research level mathematics, long horizon coding and advanced visual question answering. These are vendor statements. BenchGecko has no independent scores for Hy4 yet and will add them to the model page as they arrive.
Source: BenchGecko
Written automatically from BenchGecko's tracked data for this period; every number comes from the changelog, model, mindshare and pricing datasets. Checked against the source data before publishing.