Open-weights 295B MoE model with 21B active parameters for reasoning and agentic coding. It is close to much larger open models on benchmarks, and the FP8 version is about 300 GB.
Tencent's open-weights 295B Mixture-of-Experts model (21B active) for reasoning and agentic coding.
Updated EU status checked Last shown
Hy3 at a glance
- What is Hy3?
- Hy3 is a model from Tencent. Tencent's open-weights 295B Mixture-of-Experts model (21B active) for reasoning and agentic coding.
- Who makes Hy3?
- Hy3 is made by Tencent.
- Does Hy3 work in the EU?
- Unverified. No explicit evidence found. Weights are reported as open, but the license terms (including any territorial exclusions, as in some earlier Tencent Hunyuan licenses) were not available. Last verified 14 Nov 2026.
- How much does Hy3 cost?
- Open source. Weights reported as openly downloadable; hosted API pricing not confirmed.
- Where was Hy3 shown?
- In 1 creator video by AI Search. Most recently by AI Search on 12 Jul 2026, at 33:12 in “Girlfriend simulators, GPT 5.6, Grok 4.5, Seedream 5.0, Muse Spark, robot surgery: AI NEWS”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Hy3 (also called High3) is a large language model from Tencent's Hunyuan team. According to a YouTube creator summary, it is a Mixture-of-Experts model with about 295B total parameters and 21B active per token, aimed at reasoning and agentic coding tasks.
Reported benchmark results put it close to much larger open models. The FP8 version of the weights is described as roughly 300 GB, so self-hosting requires multi-GPU server hardware.
The official page (hy.tencent.com/research/hy3) was only available as a title in the material provided, so the license, exact release date, context length and pricing could not be confirmed here. Weights are reported to be open, but no download link was provided.
- 295B total parameters, 21B active (Mixture-of-Experts)
- Targeted at reasoning and agentic coding
- Open weights reported
- FP8 checkpoint of about 300 GB
- Benchmarks reported close to much larger open models
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Similar AI tools
Closest neighbours by meaning, not just by category.
Big Bang v1
Model
Open-weights 35B-A3B agentic model post-trained from Qwen3.6-35B-A3B for search, coding and scientific research.
LongCat 2.0
Model · Meituan
Meituan's open-weights 1.6T-parameter mixture-of-experts LLM (48B active), aimed at agentic coding and reasoning tasks.
HY-MT2
Model · Tencent
Tencent's open multilingual translation model family (1.8B, 7B, 30B-A3B MoE) covering 33 languages.
Hy4
Model · Tencent
Tencent's Hy4 multimodal model with visual understanding, computer use, 3D/Blender and knowledge-work abilities.
Agents A1
Model
35B Mixture-of-Experts open-weights model from InternScience for long-horizon agentic search, engineering and tool use.
Nanbeige4.3-33B
Model · Nanbeige
33B dense open-weights LLM from Nanbeige, tuned for agentic coding and professional cowork tasks.