FlyRouter

Guides

Models

FlyRouter offers text models from the Claude, GPT, Gemini, Grok and DeepSeek families, plus image models such as GPT Image and Nano Banana. Model IDs match each vendor's official IDs, so you don't need to rename anything when switching to FlyRouter.

List available models

The catalog changes as new models launch, so treat the API as the source of truth:

Terminal
curl https://api.flyrouter.ai/v1/models \
  -H "Authorization: Bearer $FLYROUTER_API_KEY"

This returns every model the current key can call (if the key is limited to certain models, only those are returned):

json
{
  "object": "list",
  "data": [
    { "id": "claude-sonnet-5-5", "object": "model" },
    { "id": "gpt-6.1-sol", "object": "model" }
  ]
}

Per-model prices, context length and cache prices are on the model pricing page.

Main model families

FamilyExample model IDsRecommended protocol
Claudeclaude-opus-5-5, claude-sonnet-5-5, claude-haiku-5-5, claude-fable-5-1Anthropic native
GPTgpt-6.1-sol, gpt-6-astra, gpt-6-lunaOpenAI compatible (Chat Completions or Responses)
Geminigemini-4-argonGemini native or OpenAI compatible
Grokgrok-4.7, grok-4.6OpenAI compatible
DeepSeekdeepseek-v4-pro, deepseek-v4-flashOpenAI compatible
Imagegpt-image-2, nano-banana-2, nano-banana-proOpenAI Images

Choosing a model

  • Coding and complex agent work: claude-opus-5-5, claude-sonnet-5-5 or gpt-6.1-sol. In Claude Code, use Sonnet as the workhorse and Opus for hard problems.
  • Strongest reasoning: claude-fable-5-1, for long, demanding reasoning and planning. It also has the highest price.
  • Everyday chat, support and bulk processing: lightweight models such as claude-haiku-5-5 and deepseek-v4-flash, which are fast and cheap.
  • Understanding images: the Claude, GPT and Gemini families all accept image input. See Vision.
  • Generating images: see Image generation.

Protocols and models

Every text model can be called through the OpenAI-compatible /v1/chat/completions. For Claude, though, use the Anthropic native protocol to get full prompt caching and reasoning support. /v1/responses supports GPT models only.

When a model isn't available

If you get model_not_found or a 503:

  1. Check the model ID against GET /v1/models, watching for - versus . (for example claude-sonnet-5-5 versus gpt-6.1-sol).
  2. Make sure the key isn't limited to other models.
  3. Make sure the endpoint fits the model; for example, don't call Claude through /v1/responses.
FlyRouter

One key, every top AI model