Hy3
Open Weights
Tencent
Tencent's open-weight reasoning and agent model, a 295B-parameter mixture-of-experts with 21B active parameters (plus a 3.8B MTP layer) and a 256K-token context window, released under Apache 2.0. It follows the Hy3 Preview launch in late April 2026 with scaled-up post-training. Secondary coverage often calls it 'Hunyuan 3.0', but Tencent's own model card uses only 'Hy3'.
Strengths
- Apache-2.0 licensed, one of the most permissive licences among frontier-scale open models
- 295B total parameters with only 21B active, giving strong cost efficiency
- 256K-token context window
- Scaled-up post-training over the late-April Hy3 Preview release
- Scored 2.67/4 against GLM-5.1's 2.51/4 in a blind evaluation by 270 experts
Caveats
- Text generation only - no vision or audio
- Tencent publishes most benchmark results as images, so exact scores are not recorded here
- No release date is stated on the model card
- Serving a 295B MoE still requires substantial GPU memory despite the low active count
- Vendor
- Tencent
- Context Window
- 256,000 tokens
- License
- Apache-2.0
- Modalities
- text
| Benchmark | Score | Max |
|---|---|---|
| Expert blind evaluation | 2.7 | 4.0 |
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
Open weights under Apache-2.0 - self-host at no licence cost. A 295B MoE with 21B active parameters, so inference is far cheaper than the total parameter count suggests.