TrustList summary

Not yet independently verified. Not yet independently verified. Tencent has announced open weights, but we could not find them on Hugging Face, so the licence and parameter count are unconfirmed; the context window and prices are from OpenRouter, and the evaluation scores are Tencent's own. We will update this when it can be confirmed, and remove this note.

Tencent announced Hy4 on 4 November 2026 and describes it as joining the leading open source models, with weights and system cards due "in the coming days". Until those arrive, it can be tried through hosted services such as OpenRouter, and its licence cannot be read.

Sight and computer use arrive in the Hy line

Tencent says Hy4 is the first model in the line with visual understanding, and the first able to use a computer, through both the command line and a graphical interface. It also stresses agentic work. Against Hy4 Preview, Tencent reports that the hallucination rate fell from 12.95% to 7.2%, on its internal measure.

A finance test Tencent ran itself

The company's headline comparison is a blind evaluation of financial work, using 213 financial professionals and 579 tasks. It reports Hy4 at 3.04, DeepSeek-V4.1-Flash-1017 at 2.97, and Kimi K3.1 Pro and GLM-5.3 at 2.68. Tencent ran the test, so we describe it here and do not store it as a benchmark score.

Key facts

Maker: Tencent Announced: 4 November 2026 Inputs: text and images, per OpenRouter Context window: 1,048,576 tokens, with up to 65,536 tokens of output, per OpenRouter Hosted price: $0.834 input and $2.501 output per million tokens undiscounted on OpenRouter Weights: announced, not yet found by us

Benchmarks & AI stats

AI model rankings

Benchmark	Official	Community avg
No benchmark scores yet. Be the first to add one.

“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).

Sign in to add a benchmark score for this model.