OpenRouter Now Hosts Tencent Hy3-Preview for Free Agentic Reasoning

OpenRouterOpenRouter

· Updated

OpenRouter is now hosting Tencent's new Hy3-preview model, offering free access to the 295B-parameter Mixture-of-Experts model. This integration allows developers to test frontier-level reasoning and coding capabilities with a 256K context window at no cost.

OpenRouter, a unified API platform for accessing hundreds of language models, is now hosting the Tencent Hy3-preview model for free. The platform's official handle announced the model features a 295 billion parameter Mixture-of-Experts architecture with 21 billion active parameters and a 256K context window, following Tencent's open-source release of Hy3-preview.
Model
Tencent Hy3-preview
Total parameters
295B
Active parameters
21B
Context window
256K tokens
Availability
OpenRouter
Pricing
Free through May 8, 2026

This integration provides cost-free access through May 8, 2026, to controllable reasoning effort—allowing users to toggle between direct responses and deep thinking. It extends OpenRouter's strategy of offering free windows, mirroring the integration of long-context reasoning on other platforms.

Access the model through the OpenRouter API or chat interface for free through May 8, 2026. It is suited for coding agents and complex search tasks requiring its native 256K context window. While OpenRouter provides the API, the model weights remain available on Hugging Face for local deployment.

View the full update on openrouter.ai
OpenRouter
OpenRouter
@OpenRouter
X

The new Hy3-Preview model from @TencentHunyuan is live for free on OpenRouter! It’s a 295B MoE model (21B active) with controllable reasoning effort. A cost-effective, practical model that performs strongly in coding agents and delivers comprehensive general capabilities. https://t.co/5r4UbLgxwg

7retweets159likes
View on X

Still wondering? A few quick answers below.

OpenRouter is now hosting Tencent's Hy3-preview, a 295 billion parameter Mixture-of-Experts model. It activates only 21 billion parameters during inference, providing high-efficiency reasoning and a 256,000-token context window. It is currently available for free on the platform, allowing developers to test its agentic coding and search capabilities without cost.

Users can control the reasoning effort of Hy3-preview through the OpenRouter API or chat interface. By adjusting the reasoning effort parameter, you can toggle between direct responses for simple tasks and deep chain-of-thought thinking for complex math or coding. This flexibility allows for optimizing token usage and response quality based on specific workflow requirements.

Yes, OpenRouter is offering free access to the Hy3-preview model through May 8, 2026. This promotional window allows developers and researchers to integrate the model into agentic workflows and test its performance on long-context tasks without incurring API costs. After this period, the model will likely transition to a standard paid tier on the platform.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update