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
| Protocol | Endpoint | Best for | Request shape | Official docs | OmniRouters note |
|---|---|---|---|---|---|
| OpenAI Chat Completions | /v1/chat/completions | Best default for most projects | messages array | OpenAI Chat | Safest universal option across model families |
| OpenAI Responses | /v1/responses | Newer structured workflows | OmniRouters currently documents a messages-based schema | OpenAI Responses | Check Apifox for exact field behavior |
| Claude Messages | /v1/messages | Anthropic SDK migration | Claude-style messages, max_tokens, system | Anthropic Messages | Uses bearer auth on OmniRouters, plus anthropic-version |
| Gemini Generate Content | /v1beta/models/{model}:generateContent | Gemini SDK migration | contents and parts | Gemini generateContent | Uses bearer auth on OmniRouters; multimodal uploads should follow OmniRouters constraints |
Practical Guidance
Start here for new builds
- Choose OpenAI Chat Completions if you want the cleanest default
- Choose OpenAI Responses if you are designing around structured outputs or tool workflows
Start here for migration projects
- Choose Claude Messages if you already ship Anthropic payloads
- Choose Gemini Generate Content if you already ship Gemini payloads
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.