Hy3 AI Model: Features, Context and Pricing
tencent/hy3Hy3 is Tencent’s open-weight text model for reasoning, tool use, and agent workflows.
262K tokens of context — about 196,608 words, on a scale that ends at 2M.
What it can do
- Reasoning
- Tool use
- Open weights
- Takes text
- Gives text
Released 6 Jul 2026 · checked 3 Oct 2026.
About it
Hy3 is a Tencent language model with 295 billion total parameters and 21 billion active at a time. It handles text input and output, supports reasoning and tools, and can read up to 262,144 tokens in one context. Open weights are available, and pricing is $0.0825 per million input tokens and $0.33 per million output tokens.
Review
In a line
Hy3 is Tencent’s open-weight language model for text-based reasoning and agent workflows. It uses a Mixture-of-Experts design, with 21 billion of its 295 billion parameters active for each task.
What it's good at
Hy3 is suited to reasoning tasks, tool-assisted work, and agent workflows that carry out several steps. It supports configurable reasoning effort, so a service can adjust how much deliberate processing it uses.
The model accepts text and produces text. It also supports tools, which lets connected software call functions during a task.
Price in plain terms
Hy3 costs $0.0825 per million input tokens and $0.33 per million output tokens. Input tokens are the text you send; output tokens are the text it writes.
At the stated rates, reading a 500-word page costs under 0.01 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.025 cents. 1,000 such turns cost about $0.25.
Context and limits
Hy3 can take up to 262,144 tokens in one request, so a single task can include a large amount of text. It can generate up to 128,000 tokens in one answer.
Its input and output are text only. The facts describe reasoning and tool support, but do not list image handling or other media modes.
Ways to use it
Use Hy3 in a service that exposes Tencent’s model, or connect it through a provider’s API when available. The model ID is tencent/hy3, and its open weights are listed as tencent/Hy3.
Text goes in and text comes back. Tool support can let an attached application run actions while Hy3 works through a task.
Verdict
Hy3 fits people who need long-context text reasoning, multi-step agent work, or tool calls and want an open-weight Tencent model. Its low token rates also suit applications with many small conversations.
Choose another model if you need image understanding or image creation, since Hy3 is text-only. You also need a service or setup that can host or expose its open weights; the facts do not describe a bundled chat app.
Verdict
Hy3 is a practical option for teams building text agents that need long prompts, reasoning controls, and tool calls. Its open weights give providers room to host or adapt the model, while the token rates keep high-volume text use inexpensive. The trade-off is scope: it handles text rather than images, and using it depends on finding a service or API setup that exposes the model. It makes the most sense when those requirements match your project, not when you want a ready-made multimodal assistant.
- Open weights are available.
- Supports reasoning with configurable effort.
- Tool support suits multi-step agent tasks.
- 262,144-token context window.
- Low stated input and output token prices.
- Text input and output only.
- Requires a service or setup that exposes the model.
- Large 295B-parameter design may be demanding to host.
- Pick it ifText agents with long prompts; Tool-using workflows; High-volume, low-cost text processing
- Skip it ifYou need image input or image output.; You want a bundled chat app rather than a model service.; You cannot host or access open model weights.
Pricing
Usage is metered by tokens: $0.0825 per million tokens sent and $0.33 per million tokens generated. The stated examples put 1,000 typical chat turns at about $0.25.
Questions
What is Hy3?
Hy3 is Tencent’s 295-billion-parameter Mixture-of-Experts language model, with 21 billion parameters active for a task.
Is Hy3 open source?
Its weights are listed as open. The model identifier is tencent/Hy3.
How much does Hy3 cost?
Input costs $0.0825 per million tokens. Output costs $0.33 per million tokens.
How much text can Hy3 read?
Its context window is 262,144 tokens, meaning that is the maximum text it can take in one request.
How long can a Hy3 answer be?
The maximum output is 128,000 tokens.
Can Hy3 use tools?
Yes. Tool support lets connected software provide functions for the model to call during a task.
Does Hy3 handle images?
No image capability is listed. The stated inputs and outputs are text.