Last known major update
November 11, 2026
bf16.ai launched version 1.0 of its API.
Official announcementAbout the tool
bf16.ai hosts open-weight language models with their refusal behavior removed and serves them through an API. The current lineup is Uncensored DeepSeek V4.1 Flash (also in a low-latency Fast version), GLM-5.3 Flash, GLM-5.3 and Kimi K3.1 Pro, all with a 1M token context window, and most of them read images as well as text. The company says the change is made in the model weights with an orthogonal projection that removes one refusal direction, with no retraining and no prompt tricks. The API speaks the OpenAI chat completions and responses formats and the Anthropic messages format, so one key works with the OpenAI and Anthropic SDKs and with coding agents such as Claude Code, Codex, OpenCode, Kilo Code, Zed, Goose and Qwen Code. The site gives ready commands to point a single Claude Code session at bf16.ai while other terminals stay on a normal plan, and setup steps for personal agents like OpenClaw, Hermes Agent and OpenMausBot. Listed uses are security audits of your own code and servers, fiction and game writing, and clinical or academic research. Prompts and outputs are not stored; only billing data such as model, token counts and cost is logged. A browser Playground lets you test each model. There is no subscription. New accounts get $0.20 of free credit for 30 days with no card, and after that you top up a prepaid USD balance with $10 to $2,000. Input costs $0.26 to $2.90 per 1M tokens and output $1.00 to $9.90 per 1M tokens, depending on the model. Purchased credit never expires, and unspent purchased credit can be refunded within 14 days of payment. Higher top-up totals unlock higher rate limits, and dedicated H200 or B200 GPU pods are available on request. Accounts are for users aged 18 and over.
Key use cases
Building AI-native applications
Create your own AI products, from copilots and assistants to intelligent workflows and dynamic interfaces.
Task Automation
Trigger actions from events automatically, with less manual coordination in project work and service delivery.
API & developer readiness
OpenAI-compatible endpoints POST /v1/chat/completions and POST /v1/responses, and an Anthropic-compatible POST /v1/messages endpoint with token counting, at https://api.bf16.ai, plus a remote MCP endpoint. One API key works across the OpenAI SDK, Anthropic SDK, Claude Code, Codex, OpenCode and OpenClaw. Keys support daily and total USD spending caps. Docs at https://bf16.ai/docs.


