Scoutari

AI builder 的決策情報台。

瀏覽

  • 今日情報
  • Skill 榜單
  • AI 週報
  • 關於與編輯方針
  • RSS

主題

  • 模型發布
  • API 與價格
  • MCP 與 Skills
  • AI Coding
  • 中國模型
  • 開源模型
  • Agent
  • Infra / 成本
  • 值得試用
© 2026 Scoutari · scoutari.com摘要由 AI 輔助生成,並始終附上原文連結。
情報模型收藏週報設定
Tuesday, July 7, 2026Scoutari
Log inSign up
Sign up
ModelsSkill RadarSavedPersonalizeSubmit sourceLog in
IntelModelsSkill RadarWeeklySavedPersonalizeSubmit source
Back to timeline

Tue, July 713:59Open SourceModel releasesAPI & pricingOpen source

Tencent open-sources 295B MoE model Hy3, only 21B active params, 256K context

View original

Decision Brief

What changedTencent Hy team open-sources 295B MoE model Hy3, activating only 21B params per token with a 256K context window.
Why it mattersHy3 scores 78.0 on SWE-Bench Verified, Apache 2.0 licensed, ideal for low-cost local inference or agent tasks.
Who should careOpen-source model users
Affected stackNo specific stack identified
Source confidenceMedium · Reliable media or first-hand reporting

Tencent Hy team launched Hy3, a 295B MoE model activating only 21B params per inference, reducing compute. Apache 2.0, 256K token context, optimized for reasoning, agents, and long texts. Achieves 78.0 on SWE-Bench Verified, claims lower hallucination. Free trial on OpenRouter until July 21, 2026. For teams using open-source models, Hy3's 21B effective params means latency and memory near medium models but knowledge capacity of a 300B model. 256K context suits code or long docs. API devs can test free via OpenRouter; private deployment gets flexible commercial rights under Apache 2.0.

Summary basis: official / RSS sourceCompiled from the source scope noted above; the original remains authoritative.

Sources

  • MarkTechPost

    Fast research-paper and ML tooling summaries, useful for infra and agent updates.

  • MarkTechPost

Related intel

  • Gemini CLI Open-Sourced: Use Gemini Models in Terminal

    Google open-sources gemini-cli, enabling developers to interact with Gemini models directly from the command line.

  • FunASR Industrial Speech Recognition Toolkit: 170x Real-Time, 50+ Languages, Speaker Diarization & Emotion Detection

    FunASR is an industrial-grade open-source speech recognition toolkit supporting 170x real-time factor, 50+ languages, speaker diarization, emotion detection, streaming, and OpenAI-compatible API.

  • Graphify: Convert Code & Docs into Queryable Knowledge Graphs

    Graphify-Labs/graphify lets users transform any folder (code, SQL, docs, images, videos) into a queryable knowledge graph, supporting multiple AI coding assistants.

  • AI-Research-SKILLs: Open-Source AI Research & Engineering Skill Library for Agents Like Claude Code

    Orchestra Research releases AI-Research-SKILLs, an open-source skill library for any AI model, turning agents like Claude Code, Codex, and Gemini into full-fledged research & engineering assistants.