Tencent releases Hunyuan 3 Preview, a 295-billion parameter open-source Mixture-of-Experts model.

// 1h agoMODEL RELEASE

Tencent releases Hunyuan 3 Preview, a 295-billion parameter open-source Mixture-of-Experts model.

Tencent has open-sourced Hunyuan 3 Preview, a massive 295-billion parameter Mixture-of-Experts (MoE) model with an expansive 256K token context window. This model is built on an overhauled training infrastructure and introduces a hybrid fast and slow thinking approach, optimizing its performance for complex reasoning, instruction following, and advanced software engineering use cases.

// ANALYSIS

Tencent's release of such a large open-source model signals a strong push into the competitive landscape of massive open-weight LLMs.

–The 295B parameter count makes it one of the larger open-source models available, requiring substantial hardware for deployment and fine-tuning.
–The 256K context window positions it well for enterprise use cases like long document analysis and complex coding tasks.
–Its hybrid thinking methodology indicates an architectural focus on dynamic compute allocation to improve complex reasoning while maintaining efficiency.

// TAGS

tencenthunyuanaillmopen-sourcemoe

DISCOVERED

1h ago

2026-06-01

PUBLISHED

1h ago

2026-06-01

RELEVANCE

9/ 10

AUTHOR

Rob The AI Guy

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE15m ago

agent-browser CLI measures Core Web Vitals

Chris Tate has announced that agent-browser, an open-source headless browser automation CLI for AI agents, now supports measuring Core Web Vitals directly. Using the new `vitals` command, developers and agents can capture performance metrics like LCP, CLS, and React hydration details in an automated pipeline.

INFRA20m ago

DigitalOcean releases Serverless Inference engineering deep dive

DigitalOcean has published a detailed breakdown explaining how their Serverless Inference works under the hood. The deep dive covers how they handle request routing and ensure reliable, scalable AI model responses without excessive costs, addressing common challenges teams face when deploying models.

VIDEO30m ago

Mastra hosts a featured discussion with Google DeepMind's Ivan Leo on building advanced agentic workflows using Anthropic's Claude Opus 4.8 and the Mastra Agent Builder.

Mastra has highlighted a detailed session featuring Google DeepMind AI engineer Ivan Leo to discuss building advanced agentic workflows using Anthropic's Claude Opus 4.8. The discussion showcases how developers can leverage the Mastra Agent Builder, an enterprise-grade visual studio that allows teams to safely build, chat with, and publish AI agents in a secure TypeScript environment.