T
Hy3OPEN
Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use.
Prices update automatically — checked daily against provider list prices.
- $0.14 per 1M tokens, assuming 3 input for 1 output the convention used by Artificial Analysis
- cheaper than 69% of the 191 open-weight models tracked here
- 4 of 6 Tencent models tracked here, from cheapest
Benchmarks & performance
Independent benchmark scores for Hy3, measured by Artificial Analysis. Higher is better.
For the same budget, Hy3 delivers 28 times more capability. Open weights also mean you can self-host it and pay nothing per token.
Models at this level cost $1.69 per 1M tokens (median of 16) — this one costs $0.14.
Cheaper and better on this index: GLM 5.3 Flash · DeepSeek V4 Flash 0731 · DeepSeek V4 Flash 0423
About this model
Hy3 is an open-weight AI model by Tencent. You can download and self-host it for free; the prices below are hosted-API list prices, tracked daily, for when you prefer convenience over self-hosting.
Frequently asked questions
What is Hy3?
Hy3 is an AI language model from Tencent. It is open-weight: you can download it and run it on your own hardware, for free. It scores 42.2 on the Artificial Analysis intelligence index.
Is Hy3 free?
The weights are free and open — you can self-host Hy3 and pay nothing per token. If you prefer a hosted API, list prices are $0.08 per million input tokens and $0.33 per million output tokens.
What is Hy3 good at?
Independent benchmarks from Artificial Analysis give it GPQA 89.7%, Humanity's Last Exam 33.5%, Long Context Reasoning 74.7%, SciCode 47.6%, τ-Bench Banking 22.9%, Terminal-Bench 64.4%. It is particularly used for code generation.
How fast is Hy3?
It generates about 70.5 tokens per second, with a median 1.8s delay before the first token. Measured independently by Artificial Analysis.
Can I self-host Hy3?
Yes. Hy3 has open weights, so you can download it and run it on your own GPU or server with tools like Ollama, vLLM or llama.cpp — with no per-token cost.