{
  "schema_version": "rangoon.article.v1",
  "id": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/",
  "url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/",
  "markdown_url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/index.md",
  "json_url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/index.json",
  "title": "Tencent released Hunyuan-A13B under a territory-limited community license, separating open weights from universal permission",
  "summary": "Tencent’s June 27, 2025 Hunyuan-A13B release paired 80B-total and 13B-active MoE claims with a territory-limited community license.",
  "language": "en",
  "author": {
    "name": "Rangoon Editorial",
    "url": "https://rangoon.ai/insights/editorial/"
  },
  "publisher": {
    "name": "Rangoon",
    "maintainer": "Hypler",
    "url": "https://rangoon.ai/"
  },
  "format": "News analysis",
  "topic": {
    "slug": "models",
    "label": "Models",
    "url": "https://rangoon.ai/insights/topics/models/"
  },
  "tags": [
    {
      "slug": "tencent",
      "label": "Tencent",
      "url": "https://rangoon.ai/insights/tags/tencent/"
    },
    {
      "slug": "mixture-of-experts",
      "label": "mixture of experts",
      "url": "https://rangoon.ai/insights/tags/mixture-of-experts/"
    },
    {
      "slug": "hybrid-reasoning",
      "label": "hybrid reasoning",
      "url": "https://rangoon.ai/insights/tags/hybrid-reasoning/"
    },
    {
      "slug": "open-weights",
      "label": "open weights",
      "url": "https://rangoon.ai/insights/tags/open-weights/"
    }
  ],
  "dates": {
    "published": "2026-10-24",
    "updated": "2026-10-24",
    "event": "2025-06-27"
  },
  "event_source_id": 1,
  "retrospective": true,
  "reading_minutes": 3,
  "takeaways": [
    "Hunyuan-A13B’s published release date and model description were paired with a custom community license, not a general-purpose license label.",
    "Fast and slow reasoning modes are model-route choices that need separate latency, output and tool-argument tests.",
    "Tencent’s supplier benchmarks and deployment claims do not prove capability, support or legal permission in another environment."
  ],
  "source_note": "Historical reporting: Tencent’s Hunyuan-A13B community license states a June 27, 2025 release date and explicitly says it does not apply in the European Union, United Kingdom or South Korea. Tencent Open Source’s July 2 community article attributes a June 27 release of an 80B-total, 13B-active-parameter hybrid-reasoning MoE model and describes public GitHub and Hugging Face distribution. The article’s 256K, tool-use, deployment and benchmark language is Tencent reporting. “Open source” in that article does not erase the exact community-license terms or make a model output authorized to invoke a tool.",
  "sources": [
    {
      "id": 1,
      "title": "Tencent Hunyuan: Hunyuan-A13B Community License Agreement",
      "url": "https://github.com/Tencent-Hunyuan/Hunyuan-A13B/blob/main/LICENSE",
      "published": "2025-06-27"
    },
    {
      "id": 2,
      "title": "Tencent Open Source: Hunyuan-A13B hybrid-reasoning MoE release",
      "url": "https://developer.cloud.tencent.com/article/2536652?policyId=1003",
      "published": "2025-07-02"
    }
  ],
  "image": {
    "url": "https://rangoon.ai/assets/editorial/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025-diagram-v1.webp",
    "thumbnail_url": "https://rangoon.ai/assets/editorial/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025-diagram-v1-thumb.webp",
    "alt": "Original open models explanatory diagram, Mode and authority boundary. Labels: Input, Fast mode, Slow mode, Schema, Approval, Action. Source boundary: dated primary release reporting; original editorial analysis.",
    "type": "original-diagram",
    "caption": "Mode and authority boundary. Original Rangoon explanatory diagram: Source boundary: dated primary release reporting; original editorial analysis. No vendor interface, photograph, measurement result, or endorsement implied.",
    "credit": "Rangoon / Hypler",
    "source_url": "https://rangoon.ai/brand/",
    "rights_url": "https://rangoon.ai/terms/#media-use",
    "rights_note": "Hypler artwork permissions do not grant rights to third-party marks or imply endorsement."
  },
  "sections": [
    {
      "id": "release-record",
      "url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/#release-record",
      "heading": "The public record distinguishes the June 27 model release from a later explanatory post",
      "paragraphs": [
        "Tencent’s Hunyuan-A13B repository license identifies the release date as June 27, 2025. Tencent Open Source then published a Chinese-language developer-community article on July 2 describing the June 27 event, calling Hunyuan-A13B its first open-source hybrid-reasoning MoE model and stating 80B total parameters with 13B active parameters. The article says that models were available on GitHub and Hugging Face. This creates a clear historical boundary: June 27 is the artifact-release date; the July post is later first-party explanatory context. Neither record demonstrates that a specific user downloaded a checkpoint, that a conversion is equivalent to the original artifact, or that a given hardware configuration can run it with a usable response time."
      ],
      "bullets": [],
      "source_ids": [
        1,
        2
      ]
    },
    {
      "id": "license-boundary",
      "url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/#license-boundary",
      "heading": "The community license set material territorial conditions for use and derivatives",
      "paragraphs": [
        "The repository’s Tencent Hunyuan Community License says it does not apply in the European Union, United Kingdom or South Korea and is limited to a defined territory. It defines materials, model derivatives and hosted services broadly, including distribution and use through an API or other remote means. Those terms are operationally significant. A published weights repository does not by itself establish that an organization in every jurisdiction can use, host, modify, redistribute or distill the model under the same conditions. A team should preserve the exact license revision attached to the artifact, determine whether it applies to its legal entity and distribution path, and obtain qualified advice when the terms or territory matter. Technical accessibility and legal permission are separate questions."
      ],
      "bullets": [],
      "source_ids": [
        1
      ]
    },
    {
      "id": "reasoning-mode",
      "url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/#reasoning-mode",
      "heading": "The stated fast and slow modes require separate system-level evidence",
      "paragraphs": [
        "Tencent Open Source described a default slow-thinking mode and a fast/slow switch through an instruction prefix. It also described a native 256K context window, tool-use examples and model performance across several benchmarks. Those are useful statements of the publisher’s intended interfaces and test framing. They do not make extended reasoning an audit trail, prove that a tool call has correct arguments, or establish that a long input is fully used by a particular serving stack. Reasoning modes can change token use, output timing and the moment at which an application decides an answer is terminal. An operator should compare exact mode settings against permitted tasks, schema-validity checks, latency limits and a no-action fallback. A fluent chain or successful demonstration is not authorization for a hotel lookup, code execution, file write or other external effect."
      ],
      "bullets": [],
      "source_ids": [
        2
      ]
    },
    {
      "id": "deployment-boundary",
      "url": "https://rangoon.ai/insights/archive-tencent-hunyuan-a13b-open-hybrid-reasoning-june-2025/#deployment-boundary",
      "heading": "Tencent’s deployment framing still left runtime, evidence and control decisions to the adopter",
      "paragraphs": [
        "The July post says Hunyuan-A13B had been integrated with mainstream inference frameworks and offers deployment and throughput language. Those are supplier claims made around a model release. They should not be converted into a claim that one GPU, a chosen quantization, or an API integration will meet an organization’s quality or reliability target. Every deployment still has a context limit enforced by a particular runtime, an input and output serialization path, rate and memory behavior, credentials and a failure policy. The relevant record includes the source artifact, license decision, conversion or quantization, inference engine and tests for mode switching and malformed tools. The June event supports reporting an open-weight release subject to its own terms. It does not establish universal availability, a performance guarantee or a safe autonomous agent."
      ],
      "bullets": [],
      "source_ids": [
        1,
        2
      ]
    }
  ],
  "project_context": null,
  "related_links": [],
  "permissions_url": "https://rangoon.ai/permissions.txt",
  "content_hash": "sha256:87205fbcea9efc2f4240892c73bdbca6f59f6f787d9e3a780e5e7c2df242ced2"
}