Token Router

Every model, one endpoint

Anthropic-compatible and OpenAI-compatible, so your existing tools work unchanged. Bring your own provider keys and routing is free, or run on managed credits. Tokens bill at provider list price with no markup, and prompts and responses are never stored.

13
models
8
vendors
1.0M
largest context
$0.037
from, per M input

13 of 13 models · prices are provider list price per million tokens · some models cost more past a context threshold, shown under the rate

ModelVendorContextInputOutputImages
Qwen3.5 397BAlibabaopen262K$0.45$3no
Claude Haiku 4.5Anthropic200K$1$5yes
Claude Opus 4.8Anthropic200K$5$25yes
Claude Sonnet 5Anthropic200K$3$15yes
DeepSeek V4 FlashDeepSeekopen128K$0.09$0.18no
Gemini 3.1 Flash-LiteGoogle1,048K$0.3$2.5yes
Gemini 3.1 ProGoogle1,048K$2$4 past 200K$12$18 past 200Kyes
Gemini 3.6 FlashGoogle1,048K$1.5$7.5yes
Kimi K2.7 CodeMoonshotopen256K$0.74$3.5no
GPT-5.6 LunaOpenAI400K$1$6yes
GPT-5.6 TerraOpenAI400K$2.5$15yes
gpt-oss 120BOpenAI (open weights)open131K$0.037$0.17no
GLM-5.2Z.aiopen128K$0.93$3no

Then see what the spend shipped

Every request through the rail is metered per engineer at these prices, then tied back to the pull requests it produced. Routing is the door; the attribution is the product.

See live demo