Open models. One API.

Run open-weight models on our cloud and connect frontier models with your own API keys -- one account, one integration, one bill you actually control.

Terminal
curl https://api.llmstudio.dev/v1/chat/completions \
  -H "Authorization: Bearer $LLM_STUDIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "studio/qwen3-0.6b",
    "messages": [
      {"role": "user", "content": "Hello"}
    ]
  }'

Hosted inference

Open-weight models, already warm

No infrastructure to run. We host lightweight open models on our own inference workers behind one gateway, and grow the catalog over time without changing your integration.

Qwen 3 0.6B

Hosted

studio/qwen3-0.6b

Tiny, fast general-purpose chat model. Good default for low-cost hosted usage.

Gemma 3 270M

Hosted

studio/gemma3-270m

Extremely lightweight model for simple, latency-sensitive chat.

Qwen 3 Embedding 0.6B

Hosted

studio/qwen3-embedding-0.6b

Hosted embedding model for knowledge-base indexing and retrieval.

Bring your own key

Frontier models, your account

Connect your own OpenAI, Anthropic, Google, or Mistral AI key. You pay the provider directly for that usage -- we never mark it up, and your key never touches the browser after you submit it.

OpenAIAnthropicGoogleMistral AI

One integration

Hosted and BYOK behind one gateway

Every request -- hosted or BYOK -- goes through the same authenticated gateway and the same OpenAI-compatible chat completions endpoint. Switch models without switching code.

Your app
LLM Studio gateway -- auth, quotas, routing, usage
Hosted -> LLM Studio inference worker
BYOK -> provider adapter

Web app

A chat surface for your whole team

Log in, pick a model, and go -- hosted models and every provider you've connected in one picker, with conversation history, streaming, and usage details behind a subtle details menu.

llmstudio.dev/chat
Summarize this changelog for our release notes.
Here's a three-bullet summary you can paste directly into the release notes...

Developer experience

Point your existing SDK at us

The chat completions API is OpenAI-compatible. Most integrations are a two-line change: base URL and API key.

curl
curl https://api.llmstudio.dev/v1/chat/completions \
  -H "Authorization: Bearer $LLM_STUDIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "studio/qwen3-0.6b",
    "messages": [
      {"role": "user", "content": "Hello"}
    ]
  }'

Team collaboration

Built for organizations, not just accounts

Every API key, conversation, and provider credential belongs to your organization. Invite teammates with the right level of access and manage billing centrally.

Org-scoped by default

API keys, conversations, and provider credentials belong to your organization, not one person's account.

Invite teammates by role

Owners and Admins invite people as Owner, Admin, Member, or Billing, and can revoke access instantly.

Shared usage pool

One subscription and one credit pool for the whole team -- no per-seat quota juggling.

Organization API keys

Create keys scoped to the team for shared services, separate from any individual's login.

Usage controls

Metered, capped, and auditable

Hosted usage is tracked in credits, enforced against your plan's quota and concurrency limit, and logged per request -- so nothing about your bill is a surprise.

Credits, not guesswork

Every hosted request is metered in credits, weighted per model, visible in real time in Usage.

Per-plan concurrency limits

Each plan caps concurrent hosted generations, enforced server-side before a request reaches a worker.

No overage surprises

Plans without overages enabled stop cleanly at the quota; Team and Enterprise can opt into pay-as-you-go.

Usage you can audit

Every request is logged with tokens, credits, latency, and status -- never prompt or response content.

Privacy & security

Built to hold real credentials

Provider keys and platform API keys are treated as what they are: secrets, not configuration.

Encrypted at rest

Provider API keys are encrypted (AES-256-GCM) before they ever reach the database.

Hashed API keys

LLM Studio API keys are shown once at creation and stored as a salted hash, never plaintext.

No prompt storage in analytics

Usage accounting tracks token counts and latency, not your prompt or response content.

Tenant isolation

Every query is scoped to your organization; conversations, keys, and usage never cross tenants.

Pricing

Simple, config-driven plans

Every limit shown here comes straight from our plan registry -- nothing hard-coded in the UI.

Free

$0

For experimenting.

Pro

$20/mo

For builders who need more headroom.

Team

$25/mo

5-user minimum. Shared organization balance.

Documentation

Everything you need to integrate

OpenAI-compatible reference docs, authentication details, and rate-limit behavior -- kept in sync with the actual API.

Start building in a minute

Create an account, generate an API key, and call a hosted model -- no infrastructure to provision.

Start building