Chinese AI Tools
ZH

Tencent Cloud

Hy3 Preview

Hy3-preview is described by Tencent Hunyuan as a 295B-parameter Mixture-of-Experts model with 21B active parameters, 3.8B MTP layer parameters and 256K context length. The English README says it improves on complex reasoning, instruction following, context learning, coding and agent tasks, and provides vLLM and SGLang deployment examples with an OpenAI-compatible local API.

Globally availableFull English UIPublic APIFreeTrusted

Quick answers

At a glance

Overview
Tencent Hunyuan's 295B-parameter open MoE model for reasoning, long context, coding and agent tasks.
Best fit
Teams evaluating open Chinese reasoning and agent models with long-context deployment requirements.
Trust
2/2 sources verified, recently checked · 2026-05-17
Coverage
100/100

Editorial verdict

Best for

Teams evaluating open Chinese reasoning and agent models with long-context deployment requirements.

Avoid if

Avoid it if you need a small model or a fully managed hosted API without infrastructure work.

Why it matters

Hy3-preview is the English-documented Hunyuan open LLM profile.

Pricing

Open weights are free to download; self-hosting requires large GPU infrastructure

Payment

Hugging Face, ModelScope, GitCode, Self-hosted GPU infrastructure

Commercial use

Commercial use should follow the current product, API, model license and billing terms.

Privacy

Review prompt, file, media upload, retention and training-use terms before sensitive workloads.

Use-case fit

Self-hosted reasoning model

Strong

Use it when open weights, long context and local serving are more important than hosted convenience.

Coding and agent evaluation

Strong

Tencent specifically calls out SWE-bench, Terminal-Bench, BrowseComp, WideSearch, ClawEval and WildClawBench-style scenarios.

Global user checklist

RegistrationConfirmedOpen model links are published through GitHub, Hugging Face, ModelScope and GitCode.
English UIConfirmedThe repository includes an English README.
API and docsConfirmedREADME includes OpenAI-compatible local serving examples for vLLM and SGLang.
Commercial usePartialReview the Tencent Hy Community License before commercial deployment.
Coverage · 100/100

Model names, quotas, release status, regional access and commercial terms can change quickly; recheck official sources before procurement or production use.

Pros

  • - 295B total parameters, 21B active parameters and 256K context are documented
  • - vLLM and SGLang deployment paths are documented
  • - Targets coding, search and agent benchmarks

Cons

  • - The model is large enough to require serious GPU planning
  • - License is Tencent Hy Community License, not a generic permissive license

Decision paths

mimo-v2-flash

deepseek-v4-api

qwen

Sources

Hy3-preview GitHub

other · en · verified 2026-05-17

Confirms model size, context, deployment, model links and license.

Tencent Hy Research

official · en · verified 2026-05-17

Official Hunyuan research site referenced by the open-source organization.

Reviews