BLK-AI · Free with a free account
BLK-AI: Bring Your Own Model
Your model. BLK-AI's whole toolkit. Free to connect and free to use.
BBAIS hosts only BLK-AI models. Any other model is yours to bring: your address, your key, your provider's bill. Connect your own open-weight model or API in a few clicks from the model menu and use it in BLK-AI: Chat, BLK-AI: Artifacts and BLK-AI: BLKBUILDER, with BLK-AI's memory, tools, Artifacts presets and builder working on it. Free to connect and use with a free MyBLKWorldHQ account; other BLK-AI features keep their BLKCOIN price.
Providers that need only an address and a key
OpenAI, Anthropic, Google Gemini, OpenRouter, Groq, Together AI, Fireworks AI, DeepInfra, DeepSeek, Mistral, xAI, Cerebras, NVIDIA NIM, Perplexity, Cohere, SambaNova, Nebius AI Studio, Novita AI, Hyperbolic, Ollama, Hugging Face, and any vLLM, TGI, SGLang, llama.cpp, LM Studio or Ollama server. All bring-your-own: BBAIS hosts only BLK-AI models and does not provide, host or resell these.
Your own open-weight model: where it runs
Your weights run on a GPU server you rent; BBAIS never hosts them. A model only uploaded to the Hugging Face Hub runs nowhere until a host runs it.
- Modal: your .modal.run app address; Modal proxy-auth token (Modal-Key/Modal-Secret) or your own key
- RunPod: Serverless endpoint ID, or a Pod's ID and port (proxy.runpod.net)
- Baseten: your deployed model's ID, or Baseten Model APIs
- Hugging Face Inference Endpoints: your dedicated endpoint's address and token
- Replicate: owner/name (or owner/name:version); Replicate's own predictions format
- Beam: your .app.beam.cloud address and token
- Cerebrium: project ID, function name and token
- Any GPU server: Vast.ai, Lambda, Fly.io, Koyeb, Northflank or your own machine running vLLM, TGI, SGLang, llama.cpp, LM Studio or Ollama
Cloud AI platforms
- Azure OpenAI / Microsoft Foundry: resource name, deployment name and key (api-key header); v1 API or a dated api-version
- Amazon Bedrock: region and model id, plus a Bedrock API key or an access key pair (AWS Signature V4)
- Google Vertex AI: project, location and a service-account key; a one-hour token is minted and renewed
- IBM watsonx.ai: region, project ID and IBM Cloud API key, exchanged for an IAM token
- Cloudflare Workers AI: account ID and API token
What it costs
Connecting and using your own model costs no BLKCOIN. Your provider bills your model's usage. A chat turn with a live search pays the chat-message price; Envision, image reading, BLKBUILDER hosting, domains and automations keep their own prices.
Questions
What is BLK-AI: Bring Your Own Model?
BLK-AI: Bring Your Own Model lets you connect your own model, an open-weight model you run or a hosted API with your own key, and use it inside BLK-AI: Chat, BLK-AI: Artifacts and BLK-AI: BLKBUILDER with BLK-AI's memory, platform tools, Artifacts presets and builder working on it. BBAIS hosts only BLK-AI models directly; every other model is bring-your-own. It is the BYOK/BYOM option that chat apps like TypingMind, LibreChat and Open WebUI and coding tools like Cursor and GitHub Copilot offer, built into a full AI platform.
Is bringing my own model free?
Yes. Connecting and using your own model is free with a free MyBLKWorldHQ account: your model's replies cost no BLKCOIN and there is no minimum charge. Your model's provider bills you for its usage (the tokens it reads and writes) at their own price. Other BLK-AI features keep their BLKCOIN price: a chat turn that runs a live search (Smart Search, Deep Search, Web Scrape) pays the chat-message price, and Envision images and videos, image reading when your model cannot see images, BLKBUILDER hosting, domains and automations keep their own prices.
Does BBAIS host other companies' models?
No. BBAIS hosts only BLK-AI models (BLKCODE BA1-Fast, BLKOmni Flash BA1 and the rest of the BLK-AI lineup). Any other model, from Llama, Qwen, Mistral, DeepSeek or Gemma open weights to hosted APIs such as OpenAI, Anthropic, Google Gemini, xAI Grok, Groq, Together AI, Fireworks AI or OpenRouter, is bring-your-own: your address, your key, your provider's bill. BBAIS never hosts, uploads or stores your weights.
Which models can I connect?
Any model behind the standard chat completions format (/v1/chat/completions), which covers vLLM, Text Generation Inference, SGLang, llama.cpp, LM Studio and Ollama servers and providers such as OpenAI, Google Gemini, Groq, Together AI, Fireworks AI, DeepInfra, DeepSeek, Mistral, xAI, Cerebras, OpenRouter and Ollama's cloud; Anthropic's Messages format, which BLK-AI speaks natively; models Hugging Face's inference providers already run, by their hub id with your own token; and your own open-weight model on Modal, RunPod, Baseten, a Hugging Face Inference Endpoint, Replicate, Beam, Cerebrium or any GPU server.
How do I connect my model?
Open the model menu in BLK-AI: Chat, Artifacts or BLKBUILDER and choose Connect your model. Paste your model's address and key; the model list fills itself (a single model is picked for you). Press Test & connect: BLK-AI sends a short live test, checks whether your model calls tools and reads images, and adds it to every model menu that can run it. Manage it any time in Dashboard → Your Models.
Can I use a model running on my own computer?
Yes, once it has a public https address. BLK-AI calls your model from the internet, so a local address like localhost or a private-network address cannot be reached and is refused for safety. Put the server behind a tunnel or reverse proxy and paste the public address.
Is my API key safe?
Your key is encrypted the moment it arrives, never shown again (you see only its last four characters), never sent to your browser and never written to logs. Only you can use, test or see your connected models, and removing a model deletes its key immediately.
What happens if my model fails?
BLK-AI tells you exactly why in plain words (rejected key, out of credit, rate limit, unreachable server) and never switches to a BLK-AI model behind your back, because that would spend your BLKCOIN. Pick a BLK-AI model in the model menu if you want to keep going, or fix your model in Dashboard → Your Models.
Do BLK-AI's tools and features work on my model?
Yes. Memory across your chats, BLK-AI's platform tools, Artifacts presets and revisions, and BLKBUILDER planning, building and editing all run on your model. A model that calls tools natively uses them directly; one that doesn't still uses BLK-AI's tools, described in its instructions. BLK-AI's platform rules apply to every request, whichever model answers.
Where can I use my own model?
Today in BLK-AI: Chat, BLK-AI: Artifacts and BLK-AI: BLKBUILDER. More environments are coming soon, starting with the BLK-AI: Cinema Director, BLK-AI: Co-Pilot, the BLK-AI: Accessibility chat and BLKBUILDER automations.
Which providers need only an API key?
Most of them: OpenAI, Anthropic, Google Gemini (AI Studio), OpenRouter, Groq, Together AI, Fireworks AI, DeepInfra, DeepSeek, Mistral, xAI, Cerebras, NVIDIA NIM, Perplexity, Cohere, SambaNova, Nebius AI Studio, Novita AI, Hyperbolic, Ollama and Hugging Face, plus any vLLM, TGI, SGLang, llama.cpp, LM Studio or Ollama server you run. Choose Address & key, tap the provider or paste its address, and paste your key. All of them are bring-your-own: BBAIS hosts only BLK-AI models.
Which providers need more than a key, and what do they need?
Azure OpenAI / Microsoft Foundry needs your resource name and deployment name with its key (sent in the api-key header). Amazon Bedrock needs a region and model id, plus a Bedrock API key or an access key ID and secret (BLK-AI signs every request with AWS Signature V4). Google Vertex AI needs a project, a location and a service-account key; BLK-AI mints a one-hour token from it and renews it. IBM watsonx.ai needs a region, project ID and IBM Cloud API key, exchanged for an IAM token. Cloudflare Workers AI needs your account ID and API token. Pick the provider under How do you connect? and fill its fields.
Where does my own open-weight model run?
On a GPU server you rent, never on BBAIS. Pick where it runs under How do you connect?: Modal (your .modal.run address with a Modal proxy-auth token or your own key), RunPod Serverless (endpoint ID) or a RunPod Pod (pod ID and port), Baseten (model ID), a Hugging Face Inference Endpoint (its address), Replicate (owner/name), Beam (app address), Cerebrium (project and function), or any other GPU server running vLLM, TGI, SGLang, llama.cpp, LM Studio or Ollama at a public https address (Vast.ai, Lambda, Fly.io, Koyeb, Northflank or your own machine).
Can I use a model I uploaded to Hugging Face?
Only once something is running it. Typing a hub id works for models Hugging Face's inference providers already run (they host it and bill your Hugging Face account). A model that is only uploaded, such as your own fine-tune, has no server yet: deploy it as a Hugging Face Inference Endpoint or on Modal, RunPod, Baseten, Beam or Cerebrium, then connect that host. BLK-AI tells you plainly when no provider runs the id you typed.