Tencent Releases Hy3 Preview MoE Model with 262K Context and Three Reasoning Modes
Tencent has released Hy3 Preview, a Mixture-of-Experts model offering 262,144 token context window and three configurable reasoning modes (disabled, low, high) for production agentic workflows. The model is available for free through OpenRouter.
Tencent Hy3 Preview — Quick Specs
Tencent Releases Hy3 Preview MoE Model with 262K Context and Three Reasoning Modes
Tencent has released Hy3 Preview, a Mixture-of-Experts (MoE) model designed specifically for agentic workflows and production deployment, according to the company. The model features a 262,144 token context window and is available for free through OpenRouter as of April 22, 2026.
Key Specifications
- Context Window: 262,144 tokens
- Pricing: $0 per 1M input tokens, $0 per 1M output tokens
- Architecture: Mixture-of-Experts (MoE)
- Reasoning Modes: Three configurable levels (disabled, low, high)
- Release Date: April 22, 2026
- Availability: OpenRouter platform
Configurable Reasoning System
The defining feature of Hy3 Preview is its three-tier reasoning system. Users can select between disabled, low, and high reasoning modes depending on task requirements. According to Tencent, this allows the model to balance processing speed against analytical depth for different use cases.
The model supports OpenRouter's reasoning parameter system, which exposes step-by-step thinking through a reasoning_details array in API responses. When continuing conversations, applications must preserve the complete reasoning details to maintain reasoning continuity across turns.
Production Focus
Tencent positions Hy3 Preview as optimized for multi-step, real-world workflows rather than benchmark performance. The company claims strong code generation capabilities and reliable performance in agentic scenarios where models must plan and execute sequences of actions.
Parameter count, training data cutoff date, and specific benchmark scores have not been disclosed.
Availability
The model is accessible through OpenRouter's API at no cost. OpenRouter routes requests across multiple infrastructure providers with automatic fallback to maximize uptime. Usage statistics are not yet available due to the recent release.
What This Means
Hy3 Preview represents Tencent's entry into the reasoning-capable model segment, joining competitors like OpenAI's o1 and DeepSeek-R1. The free pricing and production-focused design suggest Tencent is prioritizing adoption and real-world testing over immediate monetization. The 262K context window positions it competitively for document processing and long-form agentic tasks, though the absence of benchmark data makes direct performance comparisons difficult. The configurable reasoning modes offer a practical approach to the speed-vs-accuracy tradeoff that production applications face when deploying reasoning models.
Related Articles
Nvidia releases Nemotron 3 Ultra: 550B-parameter MoE model with 1M context window for agentic workflows
Nvidia has released Nemotron 3 Ultra, a 550-billion parameter mixture-of-experts model with 55 billion active parameters and support for up to 1 million token context windows. The model uses a hybrid Transformer-Mamba architecture and is designed specifically for long-running agentic workflows including agent orchestration, coding agents, and complex enterprise tasks.
NVIDIA Nemotron 3 Ultra launches on AWS SageMaker with 550B parameters, 1M token context window
NVIDIA Nemotron 3 Ultra is now available on Amazon SageMaker JumpStart with 550 billion total parameters and 55 billion active parameters. The model features a hybrid Transformer-Mamba Mixture-of-Experts architecture and supports context windows up to 1 million tokens, targeting agentic AI workloads.
Nvidia Releases Nemotron 3 Ultra: 550B Parameter MoE Model with 1M Token Context Window
Nvidia has released Nemotron 3 Ultra, a 550B parameter mixture-of-experts model with 55B active parameters and a 1M token context window. The model uses a hybrid Transformer-Mamba architecture and is available for free through OpenRouter, targeting agentic workflows and multi-step reasoning tasks.
NVIDIA releases Nemotron-3-Ultra: 550B parameter model with 1M token context and configurable reasoning
NVIDIA released Nemotron-3-Ultra-550B, a frontier-scale model with 550B total parameters (55B active) and up to 1M token context window. The model uses a hybrid LatentMoE architecture combining Mamba-2, MoE, and attention layers with Multi-Token Prediction, trained with NVFP4 quantization-aware methods from December 2025 to April 2026.
Comments
Loading...