Teams wanting a capable open model to self-host for agents get one under Apache 2.0, so UK use is not blocked by licence.
Kenning's take
What it does
Hy3 is Tencent's large language model for reasoning, coding and agent tasks, a mixture-of-experts with 295 billion parameters, 21 billion active and a 256K context. It follows Hy3 Preview from April 2026 and is aimed at cheap, dependable agent use. Weights are free under Apache 2.0 on Hugging Face, ModelScope and GitHub.
- 295B parameters, 21B active
- 256K token context
- Built for agents and coding
- Apache 2.0 open weights
- Runs on vLLM and SGLang
Who it suits. Developers and companies that want a capable open model to self-host for agents.
Key facts
- Category
- AI models
- Made by
- Tencent
- Based in
- China
- Released
- 6 July 2026
- How to get it
- open weights on Hugging Face, ModelScope, GitCode and CNB; Tencent API
- Works on
- open-weights, API
- In the UK
- Yes
- Free plan
- yes, open weights
Reviews and write-ups
- GIGAZINE 2026-07-07
FAQ
When was Hy3 released?
Tencent unveiled it on 6 July 2026, after a preview in April.
Is Hy3 open source?
Yes, under the Apache 2.0 licence.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.