OpenAI-compatible API
A drop-in replacement for the OpenAI SDK. Change the base URL, keep your code.
One API,
every model.
RouterBase routes LLM and multimodal calls through one OpenAI-compatible API — with weighted multi-channel routing and automatic failover built in.
Open source · Self-hostable · OpenAI SDK drop-in
Every model behind one endpoint
Point the OpenAI SDK at RouterBase and call whatever the catalog carries.
- Anthropic
- claude-haiku-4-5
- claude-opus-4-5
- claude-opus-4-6
- claude-opus-4-7
- claude-opus-4-8
- claude-sonnet-4
- claude-sonnet-4-5
- claude-sonnet-4-6
- claude-sonnet-5
- OpenAI
- Codex
- DeepSeek
- deepseek-prover-v2-671b
- deepseek-r1-0528
- deepseek-r1-distill-llama-70b
- deepseek-r1-distill-llama-8b
- deepseek-r1-turbo
- deepseek-v3-0324
- deepseek-v3-1
- deepseek-v3-1-terminus
- deepseek-v3-2
- deepseek-v3-2-exp
- deepseek-v3-turbo
- deepseek-v4-flash
- deepseek-v4-pro
- gemini-2-5-flash
- gemini-2-5-flash-lite
- gemini-2-5-pro
- gemini-3-1-flash-lite-preview
- gemini-3-1-pro-preview
- gemini-3-5-flash
- gemini-3-flash-preview
- ZAI
- glm-4-5
- glm-4-5-air
- glm-4-5v
- glm-4-6
- glm-4-6v
- glm-4-7
- glm-5
- glm-5-1
- glm-5-2
- gpt-4-1
- gpt-4-1-nano
- gpt-4o
- gpt-4o-mini
- gpt-5
- gpt-5-4
- gpt-5-4-mini
- gpt-5-5
- GPT-5.6
- gpt-5-nano
- MoonshotAI
- kimi-k2-0905
- kimi-k2-5
- kimi-k2-6
- kimi-k2-instruct
- kimi-k2-thinking
- kimi-k3
- MiniMax
- minimax-m1-80k
- minimax-m2
- minimax-m2-1
- minimax-m2-5
- minimax-m2-7
- o3
- o4-mini
- Alibaba
- qwen2-5-72b-instruct
- qwen3-235b-a22b-instruct-2507
- qwen3-5-122b-a10b
- qwen3-5-27b
- qwen3-5-35b-a3b
- qwen3-5-397b-a17b
- qwen3-5-plus
- qwen3-coder-30b-a3b-instruct
- qwen3-coder-480b-a35b-instruct
- qwen3-coder-next
- qwen3-vl-30b-a3b-instruct
- text-embedding-3-large
- gemini-2-5-flash-image
- gemini-3-1-flash-image
- gemini-3-pro-image
- gpt-image-2
- ByteDance
- seedream-3-0
- seedream-4-0
- seedream-4-5
- seedream-5-0-lite
- Kuaishou
- kling-v3-4k
- kling-v3-motion-control
- minimax-h3
- seedance-2-fast
- seedance-2
- sora-2
- sora-2-pro
- veo-3-1-novita
- veo-3-preview-novita
- Vidu
- vidu-q1
- vidu-q2-pro-fast
- vidu-q2-pro
- vidu-q2
- vidu-q2-turbo
- vidu-q3-pro-startend
- vidu-q3-pro
- vidu-q3-turbo-startend
- vidu-q3-turbo
- wan-2-7
One request, several channels
Channels are picked by weight. If the first one fails mid-request, the next takes over — you get one response either way.
In the catalog now
Everything the gateway can route to today.
Claude Opus 4.7
Anthropic Claude Opus 4.7 via RouterBase. Flagship Opus; prompt caching and tool calling.
$5 / 1M in
Claude Opus 4.8
Anthropic Claude Opus 4.8 via RouterBase. Latest flagship Opus; 1M-token context, adaptive thinking, mid-conversation system messages, prompt caching.
$5 / 1M in
Claude Sonnet 4.6
Anthropic Claude Sonnet 4.6 via RouterBase. Balanced cost/capability.
$3 / 1M in
Claude Sonnet 5
Anthropic Claude Sonnet 5 via RouterBase. Sonnet-tier flagship; 1M-token context, prompt caching, OpenAI-compatible chat completions.
$2 / 1M in
DeepSeek V3.2
DeepSeek V3.2 via RouterBase. OpenAI-compatible chat completions.
$0.27 / 1M in
DeepSeek V4 Pro
DeepSeek V4 Pro via RouterBase. OpenAI-compatible chat completions.
$1.74 / 1M in
Why RouterBase
What you'd otherwise build in-house
The parts that are tedious to get right, already done.
Weighted multi-channel routing
Several upstreams per model, each with its own weight. Cold-standby channels sit at weight 0 until they're needed.
Automatic failover
A failing channel is retried on the next one mid-request. Non-retryable errors stop immediately instead of amplifying cost.
Health probing
Channels that keep failing are disabled automatically, then probed with a real minimal call — never a free endpoint that lies when a provider is out of credit.
Pre-flight cost control
Cost is estimated and held against your balance before the upstream call. A runaway script can't push the account negative.
Per-attempt logging
Every attempt is logged with the channel it used and its attempt number — so you can see when a call was transferred.
Using RouterBase
Three steps to your first call
From key to response in a couple of minutes.
- 01
Create an API key
Sign up and generate an sk-rb-… key from the dashboard.
- 02
Point your SDK at RouterBase
The OpenAI SDK works as-is — just swap the base URL.
- 03
Call any model in the catalog
Same request shape you already use, streaming included.
Frequently asked questions
A unified gateway that puts several upstream providers behind one OpenAI-compatible endpoint, with weighted routing and automatic failover between them.
Start building
Every model in the catalog, behind one key.
Browse modelsNo credit card required to create a key.