Tencent Releases Hy3 Preview: Mixture-of-Experts Model with 262K Context and Configurable Reasoning
Tencent has released Hy3 preview, a Mixture-of-Experts model with a 262,144 token context window priced at $0.066 per million input tokens and $0.26 per million output tokens. The model features three configurable reasoning modes—disabled, low, and high—designed for agentic workflows and production environments.
Hy3 Preview — Quick Specs
Tencent Releases Hy3 Preview: Mixture-of-Experts Model with 262K Context and Configurable Reasoning
Tencent has released Hy3 preview, a Mixture-of-Experts (MoE) model with a 262,144 token context window, priced at $0.066 per million input tokens and $0.26 per million output tokens.
Key Specifications
The model supports three configurable reasoning levels: disabled, low, and high modes. According to Tencent, this allows users to balance computational speed against reasoning depth depending on task requirements.
Hy3 preview is designed specifically for agentic workflows and production environments. Tencent claims the model delivers strong code generation capabilities and reliable performance across multi-step, real-world workflows.
Reasoning Architecture
The model exposes its reasoning process through a reasoning_details array in API responses. When enabled, the model shows step-by-step thinking before producing final answers. To maintain reasoning continuity across conversation turns, developers must preserve the complete reasoning_details when passing messages back to the model.
The reasoning feature is controlled via a reasoning parameter in API requests, allowing developers to toggle between the three modes based on task complexity.
Availability
Hy3 preview is available through OpenRouter's API, which normalizes requests and responses across multiple model providers. The model was released on April 22, 2026, according to the OpenRouter model registry.
Model weights are available, though distribution details were not specified in the release information.
Pricing Context
At $0.066 per million input tokens, Hy3 preview positions itself in the lower cost tier for frontier models. The 3.9x multiplier between input and output pricing ($0.26 per million output tokens) is standard for models with generation-focused workloads.
What This Means
Tencent's entry with a configurable reasoning model signals continued competition in the agent-focused AI space. The three-tier reasoning system is a practical approach to the speed-versus-accuracy tradeoff that developers face when building production systems. The 262K context window places it in the extended-context category, though still below the 1M+ context leaders. The combination of MoE architecture, configurable reasoning, and competitive pricing makes this a relevant option for developers building multi-step agentic applications who need cost-effective inference with reasoning capabilities.
Related Articles
LG AI Research Releases K-EXAONE 2.0, a 750B-Parameter Open-Weight MoE Model with 262K Context
LG AI Research has released K-EXAONE 2.0, a 750-billion-parameter mixture-of-experts language model with 37B active parameters, a 262,144-token context window, and support for 10 languages. The model is open-weighted under Apache 2.0 and claims competitive results against Qwen3.5, GLM-5.1, and DeepSeek-V4 Pro on reasoning, coding, and long-context benchmarks.
OpenAI Halts Parts of Astra Model Development After It Hit 'Critical' Cybersecurity Threshold
OpenAI disclosed that its in-development Astra model showed cyberattack capabilities strong enough that it cannot rule out a 'Critical' risk classification. The company has paused related internal activity and added security controls under its Preparedness Framework.
Mistral's 3B-Parameter Shieldstral Matches 20B Safety Model on Text Benchmarks
Mistral's new Shieldstral, a 3-billion-parameter open-weight safety classifier, posts an 84.9% F1 score on text benchmarks—tying OpenAI's GPT-OSS-Safeguard-20B, a model roughly seven times larger. The model lets operators define safety rules at runtime using plain-language yes/no questions instead of fixed taxonomies.
Mistral AI Releases Shieldstral-1.0-3B, a 3B-Parameter Policy-Adaptive Safety Classifier
Mistral AI has released Shieldstral-1.0-3B, a compact open-weight safety classifier that evaluates text and images against natural-language policies specified at inference time. The 3B model runs on a single GPU and reports F1 scores competitive with or exceeding larger moderation models like LlamaGuard-4-12B and GPT-OSS-Safeguard-20B on multiple benchmarks.
Comments
Loading...