Skip to content

Protocol Comparison ​

This page helps you choose between the main LLM text-generation protocols available on OmniRouters.

First Recommendation ​

If you are unsure where to start, use an OpenAI-compatible request.

That is the most stable default on OmniRouters because all enabled models can be called through the OpenAI-compatible protocol.

Comparison Table ​

ProtocolEndpointBest forRequest shapeOfficial docsOmniRouters note
OpenAI Chat Completions/v1/chat/completionsBest default for most projectsmessages arrayOpenAI ChatSafest universal option across model families
OpenAI Responses/v1/responsesNewer structured workflowsOmniRouters currently documents a messages-based schemaOpenAI ResponsesCheck Apifox for exact field behavior
SystemOne Structured Decisions/v1/systemoneStructured classification, routing, and decisionsstate + questionsOmniRouters-specific protocolReturns one structured result for each question
Claude Messages/v1/messagesAnthropic SDK migrationClaude-style messages, max_tokens, systemAnthropic MessagesUses bearer auth on OmniRouters, plus anthropic-version
Gemini Generate Content/v1beta/models/{model}:generateContentGemini SDK migrationcontents and partsGemini generateContentUses bearer auth on OmniRouters; multimodal uploads should follow OmniRouters constraints

Practical Guidance ​

Start here for new builds ​

Start here for migration projects ​

Final Rule of Thumb ​

If two routes can solve the same task, prefer the OpenAI-compatible protocol unless you have a strong migration reason to keep a provider-native request shape.