B.AI announced today that it now fully supports the Responses API.

All existing models can be called uniformly through `/v1/responses`. At the same time, new MiMo, Qwen, Hunyuan, OpenAI, Gemini, and Grok series models have been added across different groups.

For developers, this means one less layer of adaptation. Previously, integrating models from different providers meant handling each provider's interface separately. Now, B.AI uses the Responses API as a unified entry point. You keep the familiar OpenAI-style calling convention—just switch the base_url to call other models.

Among the new models, MiMo comes from Xiaomi, Qwen from Alibaba, Hunyuan from Tencent, and Grok from xAI. Add OpenAI and Gemini to the mix, and you've got all the major players in one place. You can see exactly which models are available in which groups directly in the B.AI console.

Unified APIs are the trend, but the real value isn't in the moment of integration. The Responses API brings more structured tool calling and multi-turn state management—whether these capabilities behave consistently across different models is what's worth watching next.

Don't rush to celebrate. The interface may be unified, but model temperaments aren't necessarily. Run your prompts across models to see whether they hold up.