Hy3 preview
Summary
Hy3 preview is an open-weight language model from Tencent (Hunyuan), released on 20 Apr 2026. It accepts text and generates text, with a 262K-token context window and up to 236K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.18 per million input tokens and $0.60 per million output tokens (OpenRouter), across 1 provider we track.
Tencent Hy reasoning model for coding, instruction following, and agent tasks
Last verified 18 Nov 2026. Every figure on this page links to its source.
Specifications
| Developer | Tencent (Hunyuan) |
|---|---|
| Release date | 20 Apr 2026 [models.dev] |
| Status | Generally available [models.dev] |
| Weights | Open weights [Hugging Face] |
| Parameters | 299B [Hugging Face] |
| Context window | 262K (262,144 tokens) [models.dev] |
| Max output | 236K tokens [models.dev] |
| Knowledge cutoff | Jun 2025 [models.dev] |
| Input | Text [models.dev] |
| Output | Text [models.dev] |
| Reasoning mode | Yes [models.dev] |
| Tool calling | Yes [models.dev] |
| Structured output | No [models.dev] |
| Hugging Face | tencent/Hy3-preview [Hugging Face] |
Hy3 preview API pricing by provider
USD per million tokens. Sorted by input price.
| Provider | Input | Output | Cached input | Context | Source |
|---|---|---|---|---|---|
| OpenRouter tencent/hy3-preview | $0.18 | $0.60 | $0.06 | 262K | OpenRouter, 18 Nov 2026 |
Running Hy3 preview locally
Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.
| Quantization | Weights | Total, 8K context | Total, 32K context |
|---|---|---|---|
| Q4_K_M (4-bit) | 181 GB | 185 GB | 193 GB |
| Q8_0 (8-bit) | 317 GB | 321 GB | 329 GB |
| FP16 / BF16 (16-bit) | 598 GB | 601 GB | 609 GB |
At 4-bit with an 8K conversation, Hy3 preview needs about 185 GB, more than any single consumer graphics card; it needs a unified-memory machine, a multi-GPU setup or a split between graphics card and system RAM. Check your own hardware, or see Best computer for local AI on our sister site AIPCs.
Frequently asked questions
What is the context window of Hy3 preview?
Hy3 preview has a context window of 262,144 tokens (262K) and can generate up to 235,929 tokens in a single response, according to models.dev.
How much does the Hy3 preview API cost?
Hy3 preview is listed from $0.18 per million input tokens and $0.60 per million output tokens (OpenRouter). Prices last verified 18 Nov 2026.
Is Hy3 preview open source?
Hy3 preview is an open-weight model: its weights are published on Hugging Face as tencent/Hy3-preview. Check the license terms before commercial use.
When was Hy3 preview released?
Tencent (Hunyuan) released Hy3 preview on 20 Apr 2026, according to models.dev.
How much VRAM does Hy3 preview need?
About 185 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 321 GB at 8-bit. That is more than any single consumer graphics card holds. These are estimates from the model's published parameter count and architecture.
Similar models from other labs
Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.
| Model | Lab | Released | Context | Input $/1M | Output $/1M | Weights |
|---|---|---|---|---|---|---|
| Mistral Small 4 | Mistral AI | 16 Mar 2026 | 262K | $0.15 | $0.60 | Open |
| Qwen3.6 35B-A3B | Alibaba (Qwen) | 17 Apr 2026 | 262K | $0.10 | $0.90 | Open |
| Gemma 4 31B IT | Google DeepMind | 2 Apr 2026 | 262K | $0.14 | $0.40 | Open |
| Qwen3 Coder Next | Alibaba (Qwen) | 3 Feb 2026 | 262K | $0.12 | $0.80 | Open |
| NVIDIA Nemotron 3 Super 120B A12B | NVIDIA | 11 Mar 2026 | 262K | $0.07 | $0.40 | Open |
| Step 3.7 Flash | StepFun | 29 May 2026 | 256K | $0.20 | $1.15 | Open |
Change history
- Knowledge cutoff changed from none to Jun 2025