Coding models for Claude Code, Codex CLI, Cursor and Cline

The models coding agents actually run on, with the price you pay per 1M tokens on Crazyrouter and the one-line setup for each tool.

Crazyrouter is an AI API gateway that runs 25 coding and agentic models behind one key — including Claude Opus, Claude Sonnet, GPT Codex and DeepSeek V4 — at per-token prices from $0.165, and plugs into Claude Code, Codex CLI, Cursor and Cline by changing a base URL.

25
models
$0.165
from, per 1M input
45%
max below list
Try Gemini 3 Flash freeGet an API key
Input price per 1M tokens (list vs Crazyrouter), cheapest first
ListCrazyrouterGemini 3 Flash$0.500 → $0.165DeepSeek V4 Flash$0.300 → $0.230MiniMax M3$0.300 → $0.300MiMo V2.6 Pro$0.435 → $0.435DeepSeek V4 Pro$1.32 → $0.670GPT-5$1.25 → $0.813GPT-5 Codex$1.25 → $0.813GLM-5$1.00 → $1.00Gemini 3.1 Pro$2.00 → $1.10GPT-5.3 Codex$1.75 → $1.14Claude Sonnet 5$2.00 → $1.30GPT-4.1$2.00 → $1.30

13 more models in the table below

Tools these models power

Gemini 3 Flash sample
Gemini 3 Flash
DeepSeek V4 Flash sample
DeepSeek V4 Flash
MiniMax M3 sample
MiniMax M3
MiMo V2.6 Pro sample
MiMo V2.6 Pro
DeepSeek V4 Pro sample
DeepSeek V4 Pro
GPT-5 sample
GPT-5
GPT-5 Codex sample
GPT-5 Codex
GLM-5 sample
GLM-5

Coding & agentic models, by price

ModelVendorCrazyrouter price (per 1M tokens (in / out))ListContextDiscount
Gemini 3 Flash
gemini-3-flash
Google$0.165 / $1.32 per 1M tokens$0.50 / $3.001M tokens−45%Try · Compare
DeepSeek V4 Flash
deepseek-v4-flash
DeepSeek$0.23 / $0.67 per 1M tokens$0.30 / $1.201M tokens—Try · Compare
MiniMax M3
MiniMax-M3
MiniMax$0.30 / $1.20 per 1M tokens$0.30 / $1.201M tokens—Try · Compare
MiMo V2.6 Pro
mimo-v2.6-pro
Xiaomi$0.435 / $0.87 per 1M tokens$0.435 / $0.871M tokens—Try · Compare
DeepSeek V4 Pro
deepseek-v4-pro
DeepSeek$0.67 / $2.00 per 1M tokens$1.32 / $3.961M tokens—Try · Compare
GPT-5
gpt-5
OpenAI$0.813 / $6.50 per 1M tokens$1.25 / $10.00272K tokens−35%Try · Compare
GPT-5 Codex
gpt-5-codex
OpenAI$0.813 / $6.50 per 1M tokens$1.25 / $10.00272K tokens−35%Try · Compare
GLM-5
glm-5
Zhipu AI$1.00 / $3.20 per 1M tokens$1.00 / $3.20200K tokens—Try · Compare
Gemini 3.1 Pro
gemini-3.1-pro
Google$1.10 / $6.60 per 1M tokens$2.00 / $12.001M tokens−45%Try · Compare
GPT-5.3 Codex
gpt-5.3-codex
OpenAI$1.14 / $9.10 per 1M tokens$1.75 / $14.00272K tokens−35%Try · Compare
Claude Sonnet 5
claude-sonnet-5
Anthropic$1.30 / $6.50 per 1M tokens$2.00 / $10.001M tokens−35%Try · Compare
GPT-4.1
gpt-4.1
OpenAI$1.30 / $5.20 per 1M tokens$2.00 / $8.001M tokens−35%Try · Compare
GPT-5.4
gpt-5.4
OpenAI$1.63 / $9.75 per 1M tokens$2.50 / $15.001M tokens−35%Try · Compare
GPT-4o
gpt-4o
OpenAI$1.63 / $6.50 per 1M tokens$2.50 / $10.00128K tokens−35%Try · Compare
Claude Sonnet 4.6
claude-sonnet-4-6
Anthropic$1.95 / $9.75 per 1M tokens$3.00 / $15.00200K tokens−35%Try · Compare
OpenAI o3
o3
OpenAI$2.00 / $8.00 per 1M tokens$2.00 / $8.00200K tokens—Try · Compare
Qwen3.8 Max
qwen3.8-max
Alibaba Cloud$2.00 / $6.00 per 1M tokens$2.00 / $6.00992K tokens—Try · Compare
Kimi K3
kimi-k3
Moonshot AI$3.00 / $15.00 per 1M tokens$3.00 / $15.001M tokens—Try · Compare
Claude Opus 5
claude-opus-5
Anthropic$3.25 / $16.25 per 1M tokens$5.00 / $25.001M tokens−35%Try · Compare
Claude Opus 4.8
claude-opus-4-8
Anthropic$3.25 / $16.25 per 1M tokens$5.00 / $25.00200K tokens−35%Try · Compare
Claude Opus 4.7
claude-opus-4-7
Anthropic$3.25 / $16.25 per 1M tokens$5.00 / $25.00200K tokens−35%Try · Compare
GPT-5.5
gpt-5.5
OpenAI$3.25 / $19.50 per 1M tokens$5.00 / $30.001M tokens−35%Try · Compare
GPT-5.6 Sol
gpt-5.6-sol
OpenAI$3.25 / $19.50 per 1M tokens$4.00 / $20.00922K tokens−35%Try · Compare
GPT-6 Astra
gpt-6-astra
OpenAI$3.25 / $16.25 per 1M tokens$10.00 / $50.00922K tokens−35%Try · Compare
Claude Fable 5
claude-fable-5
Anthropic$6.50 / $32.50 per 1M tokens$10.00 / $50.001M tokens−35%Try · Compare

Prices come from the live billing table and already include the Crazyrouter discount; list prices are each vendor's public rate, for comparison only.

What is a coding API?

A coding model is a chat model tuned for tool use, long context and multi-step edits. Agent tools (Claude Code, Codex CLI, Cursor, Cline) call it hundreds of times per session, so output price and context window matter more than for a chat box.

How to choose

Autonomous agents, large refactors

Claude Opus 5 / 4.8, GPT-6 Astra. Highest success rate on multi-file tasks; budget for output tokens.

Daily driver

Claude Sonnet 5 / 4.6, GPT-5.3 Codex, GPT-5.5. The default most teams settle on.

Cheap and fast

DeepSeek V4 Flash, Kimi K3, GLM-5, Gemini 3 Flash. Good for completions, tests and boilerplate.

Set it up

Claude Code: ANTHROPIC_BASE_URL=https://api.crazyrouter.com. Codex CLI / Cursor / Cline: base URL https://api.crazyrouter.com/v1. Step-by-step on the integrations pages.

Same request from any agent

Swap model for any row in the table; the example uses gemini-3-flash.

cURL
curl https://api.crazyrouter.com/v1/chat/completions \
  -H "Authorization: Bearer $CRAZYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gemini-3-flash","messages":[{"role":"user","content":"Summarize this PR in three bullets."}]}'

FAQ

Does Claude Code work through Crazyrouter?

Yes. Set ANTHROPIC_BASE_URL=https://api.crazyrouter.com and ANTHROPIC_AUTH_TOKEN to your key; sub-agents and MCP tools follow the same setting.

Does Codex CLI work?

Yes — point the provider base URL in config.toml to https://api.crazyrouter.com/v1 and pick any model in this table.

Which coding model is cheapest?

Gemini 3 Flash at $0.165 per 1M input tokens as of the latest sync. For agents, compare output prices — they dominate the bill.

Is prompt caching supported?

For Claude models yes, through the native /v1/messages endpoint; cached input is billed at the cache rate shown on each model page.

Can I use one key across all these tools?

Yes. One Crazyrouter key works in every OpenAI- or Anthropic-compatible tool; usage shows in one billing view.