Anthropic Claude
Anthropic's Claude series, strong at coding, long-document understanding and agent tasks that need multi-step execution. It is the default model of Claude Code.
Choosing a group
| Group | Price |
|---|---|
claude | Pay 23% of official price |
All Claude models are in the claude group.
Recommended API
Prefer the native Claude Messages API, which gives full access to extended thinking, prompt caching, tool calls and other features. You can also call them through Chat Completions.
bash
curl https://api.codingrouter.ai/v1/messages \
-H "x-api-key: $CODINGROUTER_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5-5",
"max_tokens": 4096,
"messages": [{"role": "user", "content": "Rewrite this function as an async version: ..."}]
}'Choosing a model
| Series | Characteristics |
|---|---|
| Opus | Most capable, for complex coding and long-running tasks |
| Fable | Complex reasoning and multi-step agent tasks |
| Sonnet | Balance of speed and capability, the first choice for everyday coding |
| Haiku | Fastest and cheapest, for high concurrency and simple tasks |
Usage tips
- Always set
max_tokens; if you omit it, the gateway uses a default value. - Cache long system prompts or documents with
cache_control. On repeated calls the input is billed at the cache price. See Prompt caching. - When extended thinking is on,
max_tokensmust be greater thanbudget_tokens.
Representative models
| Model | Group | Good for |
|---|---|---|
claude-opus-5-5 | claude | Top capability, complex coding and long tasks |
claude-sonnet-5-5 | claude | First choice for everyday coding |
claude-haiku-5-5 | claude | Fast and low cost |
claude-fable-5-2 | claude | Complex reasoning and agents |
For the full list and prices see the model square.
