The Bottom Line: Claude Cowork and Code now support any LLMs such as GPT-5, Gemini, local models, or enterprise gateways, without Anthropic making a formal announcement.
Anthropic has quietly enabled support for external language models in Claude Cowork and Code. Practitioners can now integrate any LLMs directly into Claude Desktop—from proprietary models to open-source variants to local or enterprise-internal solutions.
Anthropic has activated support for third-party models in Claude Cowork and Code for Claude Desktop. Users can integrate various language models directly: GPT-5, GPT Open Source with 120 billion parameters, Gemini, open-weight models via OpenRouter, local models on their own machine, or enterprise-wide gateways such as AWS Bedrock, Google Vertex, or Anthropic Foundry.
The feature was introduced without an official announcement. There was no blog post and no update notification—only technical documentation was available. After more than 20 hours, no official statement had been made. The discovery was made by chance by the community on Hacker News and by security researchers like Paweł Huryn, who documented the quiet release.
For local or OpenRouter-connected models, LiteLLM proxy integration is required. The ability to integrate external models gives practitioners more flexibility in choosing their inference infrastructure—particularly relevant for teams wanting to use existing enterprise setups or specific open-source models.
Source: ainews-dev.lumi-systems.io · Published April 26, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.5.2.