Use Bearly's models from your own code
By Steven Van ·
The API works with the OpenAI and Anthropic SDKs on every Bearly account, billed to plan and credits, with zero-data-retention providers on every request.
Bearly's models can now be called from your own code through the Bearly API, which works with the official OpenAI and Anthropic SDKs. It is available on every Bearly account.
How it works
Create a key under Developers → API Keys, then change the SDK's base URL and API key. Requests, streaming and responses keep the format the SDK already uses, so parsing code does not change. No OpenAI, Anthropic or other provider key is needed.
- OpenAI SDK (Chat Completions and Responses): base URL https://api.bearly.ai/v1
- Anthropic SDK (Messages): base URL https://api.bearly.ai
In both cases the Bearly API key goes in as the API key. Then set model to a Bearly model such as gpt-6.1-sol or bearly-pro.
Models and APIs
At launch the API offers Bearly Fast, Pro and Ultra, GPT 6.1 Sol, GPT 6 Luna and GPT 6 Astra, Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5, Grok 4.7, GLM 5.3 and 5.3 Flash, DeepSeek V4 Pro and Flash, and MiniMax M3.
Each model uses the API its provider built for it, and Bearly does not translate between formats. GPT models, Grok and the Bearly models accept both Chat Completions and Responses (Bearly Fast is the exception, using Chat Completions only). Claude models use Anthropic's Messages API. GLM, DeepSeek and MiniMax use Chat Completions. Calling client.models.list() in either SDK shows what an account can use.
Privacy, billing and team policy
Every request goes only to providers that have agreed to zero data retention, with no setting to turn that off. Usage is billed to the Bearly account under its plan and credits, the same way chat is. If a streamed response is abandoned early, the model still finishes it and the full response is billed.
Managed teams' model policies apply to the API. An administrator can turn API access off with the Developer API access policy setting, in which case members do not see API Keys in the app and their requests return a 403 error.
Limits
- Responses API requests are sent with storage off, so previous_response_id, background mode and retrieving or deleting past responses are not available.
- Provider-hosted tools such as web search, file search, code execution, image generation and hosted MCP servers return a 400 error. Function calling with your own tools works.
- Priority processing is not available. Images, embeddings, audio and fine-tuning are not available either.
- Models can be added, replaced, restricted or removed at any time, and a removed model returns a model_not_found error.
