API gateway

One OpenAI-compatible endpoint across providers. Generate an API key, watch your usage, ship. Pay-as-you-go inference — MiMo, Gemini, MiniMax, Qwen · see pricing → open llm inference gateway Opengateway. One endpoint across providers — OpenAI-compatible and Anthropic-compatible. Generate an API key, swap the base URL (or ANTHROPIC_BASE_URL), and ship. Your usage is yours — visible in your console and in a live global view. sign in to generate a key →read the docs →see global usage →checking… curlpythonnode curl https://opengateway.gitlawb.com/v1/chat/completions \ -H "authorization: Bearer ogw_live_…" \ -H "content-type: application/json" \ -d '{ "model": "mimo-v2.5-pro", "messages": [ {"role": "user", "content": "hello, gateway"} ] }' One endpoint, many providers Point any OpenAI-compatible client at opengateway.gitlawb.com/v1 and pass the model. We route to Xiaomi MiMo, Google Gemini, MiniMax, and Qwen — secrets stay server-side. Smart routing with model: “auto” Pass model:“auto” and the gateway picks the cheapest model expected to handle the request, escalating automatically when an upstream fails. Billed at whatever model actually serves. Manage your own keys Generate gateway API keys from your console, scope usage to each one, revoke instantly. No embedded provider secrets in your OSS clients. Real-time usage, yours and global See your own token spend per model, plus the live global gateway view. Built on durable Tigris snapshots — no per-restart resets. smart routingNEW Stop picking models. Send model: “auto”. The gateway inspects each request — estimated context size, tool definitions, code content, reasoning params — and routes it to the cheapest model expected to produce a good answer. When an upstream rate-limits, 5xxs, or returns nothing, it escalates to the next model up the ladder before you ever see an error. You're billed at the rate of the model that actually served, and the x-gateway-served-model response header tells you which one that was. Picks the cheapest model expected to handle the request — scored from context size, tools, code content, and reasoning params Escalates automatically on rate limits, upstream 5xx, and empty responses — up to 3 attempts before you ever see an error Optional route hint: {"priority": "cost" | "balanced" | "quality", "max_cost_usd": 0.01} — stripped before it reaches any provider Billed at the serving model's listed rate — the x-gateway-served-model response header names who answered curl https://opengateway.gitlawb.com/v1/chat/completions \ -H "authorization: Bearer ogw_live_…" \ -H "content-type: application/json" \ -d '{ "model": "auto", "route": {"priority": "cost", "max_cost_usd": 0.01}, "messages": [{"role": "user", "content": "hello, router"}] }' # response header tells you who served it: # x-gateway-served-model: xiaomi/mimo-v2.5 pricing Pay only for what you use. Every model routable through the gateway, priced per token and deducted from your credit balance. Top up with the $10/mo premium plan. modelinput /1Mcached /1Moutput /1M Auto (smart routing)NEW picks the cheapest capable model · escalates on failure model: auto billed at the serving model's rate MiMo V2.5-Pro general large language model model: xiaomi/mimo-v2.5-pro $0.522$0.00432$1.044 MiMo V2.5 multimodal understanding model model: xiaomi/mimo-v2.5 $0.168$0.00336$0.336 Gemini 3.1 Flash Lite low-latency multimodal model · 1M context model: google/gemini-3.1-flash-lite $0.3$0.03$1.8 MiniMax M3 agentic reasoning model model: minimax/minimax-m3 $0.36$0.072$1.44 Qwen 3.7 Max flagship coding model model: qwen/qwen3.7-max $1.5$0.3$4.5 Kimi K3 frontier multimodal reasoning · 1M context model: moonshotai/kimi-k3 $3.6$0.36$18 GLM 5.2 agentic coding & reasoning · 1M context model: z-ai/glm-5.2 $1.68$0.312$5.28 Nemotron 3 Ultra (free)FREE frontier reasoning MoE · 1M context · free (rate limited) model: nvidia/nemotron-3-ultra-550b-a55b:free ⚡ free for everyone — no credits needed $0$0$0 Nemotron 3 Ultra frontier reasoning MoE · 1M context model: nvidia/nemotron-3-ultra-550b-a55b $0.72$0.24$4.32 Ling 3.0 Flash token-efficient agentic MoE · 262k context model: inclusionai/ling-3.0-flash $0.072$0.0144$0.216 Tencent HY3 multimodal model · 262k context model: tencent/hy3 $0.24$0.06$0.96 Macaron V1 Tall routed MoL personal-intelligence model · 262k context model: mindai/macaron-v1-tall $0.54$0.096$3.12 Macaron V1 Venti flagship routed MoL on GLM-5.2 · 1M context model: mindai/macaron-v1-venti $1.8$0.36$5.4 USD per 1M tokens · billed per token from your credit balance · cached input is reused-prompt reads. passing the model id Send the model string exactly as listed above — full provider/model ids (MiMo also accepts the short form mimo-v2.5-pro). Sending auto (or gitlawb/auto) enables smart routing — the x-gateway-served-model response header names the model you were billed for. Per-model capabilities, context windows, and copy-paste snippets live in the model catalog → what you can do A console for your own inference traffic. The gateway already routes traffic for openclaude and the rest of the gitlawb stack. The console is where you manage your own keys and see what you've spent — without leaving a tab to read logs. Use openclaude, opencode, or any OpenAI-compatible client without embedding provider keys Generate a key per project, environment, or teammate — revoke when rotated Watch personal and global token usage update in near real time Pay-as-you-go per token — top up credits and only spend what you use next step Generate your first key. Sign in with Google, X, or an email link, create a key in seconds, top up credits, and start sending requests. Usage is metered per token and billed from your balance. open console →global usage →