Chat & reasoning models — one API key, 40 models
Claude, GPT, Gemini, DeepSeek, Kimi, GLM, Qwen and Grok through the same OpenAI-compatible endpoint. Compare per-token prices side by side, then switch models by changing one string.
Crazyrouter is an AI API gateway that serves 40 chat and reasoning models through one OpenAI-compatible key, with per-token prices from $0.055 and native Anthropic, Gemini and Responses endpoints where the vendor has them.
27 more models in the table below
What the models look like in use
All chat & reasoning models, by price
| Model | Arena score | Official price | Site discount | Release date |
|---|---|---|---|---|
| Kimi K3.1 Pro | Not matched | $2.00 / $10.00 | — | |
| Kimi K3.1 Pro Highspeed | Not matched | $4.00 / $20.00 | — | |
| Claude Haiku 5.5 | 1424.8Evaluation details
| $0.10 / $0.50 | — | |
| GPT-6.1 Sol | 1481.3Evaluation details
| $2.00 / $10.00 | — | |
| Claude Sonnet 5.5 | 1477.1Evaluation details
| $2.00 / $10.00 | — | |
| GPT-6 Astra | 1475.6Evaluation details
| $10.00 / $50.00 | −35% | |
| Gemini 3.8 Flash | 1495.9Evaluation details
| $0.75 / $3.75 | — | |
| Grok 4.6 | 1455.5Evaluation details
| $2.00 / $6.00 | −15% | |
| Claude Opus 5 | 1491.6Evaluation details
| $5.00 / $25.00 | −35% | |
| Kimi K3 | 1488.2Evaluation details
| $3.00 / $15.00 | — | |
| GPT-5.6 Sol | 1484.9Evaluation details
| $4.00 / $20.00 | −35% | |
| Claude Sonnet 5 | 1461.7Evaluation details
| $2.00 / $10.00 | −35% | |
| Claude Fable 5 | 1503.9Evaluation details
| $10.00 / $50.00 | −35% | |
| MiniMax M3 | 1439.2Evaluation details
| $0.30 / $1.20 | — | |
| Claude Opus 4.8 | 1474.4Evaluation details
| $5.00 / $25.00 | −35% | |
| GPT-5.5 | 1476.7Evaluation details
| $5.00 / $30.00 | −35% | |
| DeepSeek V4 Pro | 1457.8Evaluation details
| $1.32 / $3.96 | — | |
| DeepSeek V4 Flash | 1435.7Evaluation details
| $0.30 / $1.20 | — | |
| Claude Opus 4.7 | 1493.7Evaluation details
| $5.00 / $25.00 | −35% | |
| GPT-5.4 mini | 1447.2Evaluation details
| $0.75 / $4.50 | −20% | |
| GPT-5.4 | 1465.0Evaluation details
| $2.50 / $15.00 | −20% | |
| GPT-5.3 Codex | Not matched | $1.75 / $14.00 | −20% | |
| Gemini 3.1 Pro | Not matched | $2.00 / $12.00 | −45% | |
| Claude Sonnet 4.6 | 1472.0Evaluation details
| $3.00 / $15.00 | −35% | |
| GLM-5 | 1457.2Evaluation details
| $1.00 / $3.20 | — | |
| Kimi K2.5 | 1449.7Evaluation details
| $0.60 / $3.00 | — | |
| Gemini 3 Flash | 1472.7Evaluation details
| $0.50 / $3.00 | −45% | |
| Claude Haiku 4.5 | 1414.5Evaluation details
| $1.00 / $5.00 | −35% | |
| GPT-5 Codex | Not matched | $1.25 / $10.00 | −20% | |
| GPT-5 | 1434.7Evaluation details
| $1.25 / $10.00 | −20% | |
| Gemini 2.5 Flash-Lite | Not matched | $0.10 / $0.40 | −45% | |
| Gemini 2.5 Flash | 1409.2Evaluation details
| $0.30 / $2.50 | −45% | |
| OpenAI o3 | 1431.7Evaluation details
| $2.00 / $8.00 | Not verified | |
| GPT-4.1 | 1414.6Evaluation details
| $2.00 / $8.00 | −20% | |
| GPT-4.1 mini | 1382.8Evaluation details
| $0.40 / $1.60 | −20% | |
| Gemini 2.5 Pro | 1445.4Evaluation details
| $1.25 / $10.00 | −45% | |
| GPT-4o mini | 1317.6Evaluation details
| $0.15 / $0.60 | −35% | |
| GPT-4o | Not matched | $2.50 / $10.00 | −20% | |
| Qwen3.8 Max | 1482.4Evaluation details
| $2.00 / $6.00 | — | Not verified |
| MiMo V2.6 Pro | 1479.7Evaluation details
| $0.435 / $0.87 | — | Not verified |
Text prices: USD per 1M tokens (input / output), standard base tier. Image examples show output cost; input may be billed separately. Site discounts come from the current billing configuration. Dates cite official announcements or API release notes, including previews; they are not listing dates on this site.
Leaderboard: Text · style controlled · Published:
Official Arena data; selected models and scores rounded to one decimal. Higher is better within the same leaderboard and publication date; scores are not percentages or comparable across tasks. Evaluation settings may differ from your API requests. Unmatched versions are left blank.
Source: Arena · License: CC BY 4.0
What is a chat & reasoning API?
An LLM API returns text (and increasingly tool calls and structured output) from a prompt. The models differ in reasoning depth, context window, speed and price; a gateway lets you route each request to the model that fits instead of integrating four vendors.
How to choose
Claude Opus 5 / Opus 4.8, GPT-6 Astra, GPT-5.5, Gemini 3.1 Pro. Highest output price; reserve for tasks where quality changes the result.
Claude Sonnet 5 / 4.6, GPT-5.6 Sol, Gemini 3 Flash, DeepSeek V4 Flash, Kimi K3. Best price-to-quality for summaries, drafting, extraction.
Claude Haiku 4.5, GPT-5.4 mini, Gemini 2.5 Flash-Lite, GLM-5. Fractions of a cent per request; pair with caching.
Gemini 3.1 Pro and GPT-4.1 (1M tokens), Claude (200K). Check the context column before loading a whole repo.
One request shape for every model
Open a model's detail page to copy its API model ID.
curl https://api.crazyrouter.com/v1/chat/completions \
-H "Authorization: Bearer $CRAZYROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gemini-2.5-flash-lite","messages":[{"role":"user","content":"Summarize this PR in three bullets."}]}'FAQ
What base URL do I use for chat models?
https://api.crazyrouter.com/v1 for OpenAI-compatible chat completions; Claude models also accept /v1/messages and Gemini models /v1beta.
Is the price the same as the vendor's?
No — the table shows the Crazyrouter price, which already includes the discount against each vendor's list price; the list price is shown for comparison.
Do streaming, tool calling and JSON mode work?
Yes, they behave as the upstream model documents them. Server-sent events are on by default when you pass stream: true.
Can I try a model before paying?
Yes — the newest chat models run once free with no account, from the Try-it box on their model page (no card). Every other model runs in the Playground after signup, which includes $2 of free credit; the API itself is pay as you go.
Which chat model is the cheapest on Crazyrouter?
Gemini 2.5 Flash-Lite at $0.055 per 1M input tokens as of the latest pricing sync; see the table for output prices.
Guides and comparisons
Tutorial/v1/chat/completions vs /v1/responses vs /v1/messages: Which AI API Endpoint Should You Use?
A practical guide to choosing the correct AI API endpoint. Learn the differences between OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages to avoid model unavailabl
TipsAuthorize LLM Tool Calls: Tenant Isolation with a Tested Python Executor
Validate tool names and arguments, then authorize records using trusted tenant context. One live invoice lookup and six local fixtures demonstrate the boundary.
GuideDeepSeek R2: The 32B Reasoning Model That Runs on a Single GPU — Complete Guide for Developers
DeepSeek R2 is a 32B open-weight reasoning model scoring 92.7% on AIME 2025, running on a single RTX 4090, and costing 70% less than GPT-5. Here's everything developers need to kno
Podcast
AI Dev Tools: The Crazyrouter Podcast
Short episodes for developers on AI APIs, gateways, cost control and coding tools.
Latest episodes
EP196 · Oct 29, 2026 · 6:53
AI API Security Posture — Turn Controls Into Continuous Evidence
A practical guide to AI API security posture: inventory exposure, verify controls continuously, protect keys and data, constrain tools, and turn security assumptions into evidence that operators can act on.
EP195 · Oct 28, 2026 · 7:21
AI API Policy Drift — Detect When Production Rules Stop Matching Intent
A practical guide to policy drift in AI APIs: version routing and safety rules, detect stale enforcement, compare intended and observed behavior, and restore control before small gaps become incidents.








