Computing Life · Share · YageApr 25, 2026, 08:00 UTC+8

Anthropic lets Claude Cowork run rival models, a stranger move than it looks

Anthropic 让 Claude Cowork 跑别家模型，这件事比看上去更反常

Anthropic added an April 22–23 Claude Cowork switch for GPT-5.5, Gemini 3.1 Pro, DeepSeek V4, or local models. The post says third-party deployments have no Anthropic seat fee, and Bedrock, Vertex, and gateway prompts stay outside Anthropic. The key fight is runtime and control plane: AWS, Google, and Microsoft bet on Agent Registry, Apigee, and Entra Agent ID.

Why it matters: All three HKR axes pass: the competitor-model switch is a strong hook, and the article gives billing and data-flow details. Capped below P1 because sourcing is unofficial, with no independent benchmark and a small Cowork base.

Read the original ↗Export Markdown

Take

Anthropic is taking a hard swing here: it is giving up seat revenue and prompt data on third-party model paths to keep Cowork as the enterprise agent client. The concrete hook is strong: the April 22–23 switch supports GPT-5.5, Gemini 3.1 Pro, DeepSeek V4, and local models. On Bedrock, Vertex, and gateway routes, prompts and completions do not pass through Anthropic. Telemetry can be shut off through MDM.

I do not read this as a model lab conceding weakness. In March, Anthropic blocked third-party clients from using Claude subscription tokens, then added client identity checks, fake tools, and inference signatures to Claude Code. In April, it let its own client run rival models. That pairing is the strategy. Anthropic wants stickiness to move from Claude the model to Cowork and Claude Code the runtime. AWS Agent Registry, Google Apigee, and Microsoft Entra Agent ID are fighting for the control plane; Anthropic is trying to own the layer users actually touch.

Source: Computing Life · Share · Yage · yage.ai
Published: Apr 25, 2026, 08:00 UTC+8 · Apr 25, 2026