Intel / article

Tencent Launches Hy4 Joining The Ranks Of The Worlds Leading Open Source Ai Models

Source: tencent.com
Author: Tencent editorial sitemap
Date: Published Nov 4, 2026

Context

Tencent's announcement of Hy4, its latest-generation , places it among the world's leading open-source models, but the open-source release has not happened yet. Tencent says it plans to publish the model weights, the files that let anyone run or modify a model on their own hardware, along with system cards "in the coming days". Until then Hy4 is available as a service, in Tencent products such as WorkBuddy, CodeBuddy and Yuanbao and through Tencent Cloud TokenHub, and Tencent says its API is being made available on third-party platforms such as OpenRouter. The model that can be downloaded today is its predecessor, Hy4 Preview, which Tencent published on Hugging Face in August under the Apache 2.0 license.

By Tencent's account, Hy4 adds two abilities the preview lacked. It can read images, including charts and mathematical diagrams, and it can use a computer, operating command-line and graphical interfaces to carry out tasks, with specialized training on professional software such as Blender and CAD tools. The price is unchanged from the preview, about US$0.90 per million input and US$2.70 per million output tokens.

The case for the leading-model claim rests on Tencent's own testing. In a blind evaluation run by the Tencent Hy team with Tencent Financial Technology, 213 financial professionals assessed performance on 579 real financial tasks, and Hy4 received an overall score of 3.04, against 2.97 for DeepSeek-V4.1-Flash-1017 and 2.68 each for Kimi K3.1 Pro and GLM-5.3. That head-to-head comparison covers financial work only, and most of the other figures in the announcement, including a rate that fell from 12.95% to 7.2% and gains on game-development benchmarks, compare Hy4 with Hy4 Preview rather than with competing models.

Key takeaways · AI-distilled

Hy4 adds visual understanding and computer use for the first time, operating both command-line and graphical interfaces, with specialized training on professional software such as Blender and CAD; Tencent says weights and system cards follow in the coming days.
In a Tencent-run blind evaluation of 579 real financial tasks graded by 213 professionals, Hy4 scored 3.04, ahead of DeepSeek-V4.1-Flash-1017 (2.97), Kimi K3.1 Pro (2.68) and GLM-5.3 (2.68).
Tencent says Hy4's training data excluded offensive exploitation material, and its Xuanwu and Keen labs found no significant offensive cyber concerns in the areas tested; it also rates Hy4's open-ended biology experiment design below frontier models.
Hy4 keeps Hy4 Preview's pricing of about $0.90 per million input tokens and $2.70 per million output tokens, is available through OpenRouter among other platforms, and is 50% off on Tencent Cloud TokenHub through December 31, 2026.

Terms in this piece · Glossary

LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.

Read the source www.tencent.com