Cost Estimator
Pricing updated daily · Last refreshed Oct 15, 2026Enter your per-request token volume and monthly request count to see what each model would actually cost you.
Input cost / mo
$0.7500
Output cost / mo
$0.7500
Total monthly
$1.50
on Command R7B (12-2024) — the cheapest option here
| Model | Provider | Per request | Monthly total |
|---|---|---|---|
| Command R7B (12-2024) | Cohere | $0.000150 | $1.50 |
| Llama 3.2 1B Instruct | Meta | $0.000155 | $1.55 |
| Claude Haiku 5.5 | Anthropic | $0.000450 | $4.50 |
| Llama Guard 4 12B | Meta | $0.000450 | $4.50 |
| GPT-6 Luna | OpenAI | $0.000450 | $4.50 |
| GPT-6 Luna Pro | OpenAI | $0.000450 | $4.50 |
| Ministral 3 14B 2512 | Mistral | $0.000500 | $5.00 |
| Command R (08-2024) | Cohere | $0.000600 | $6.00 |
| Mistral Small 4 | Mistral | $0.000600 | $6.00 |
| Llama 4 Scout | Meta | $0.000655 | $6.55 |
| Llama 4 Maverick | Meta | $0.000701 | $7.01 |
| Command A+ | Cohere | $0.0014 | $13.50 |
| Llama 3.3 70B Instruct | Meta | $0.0016 | $15.75 |
| Devstral 2 2512 | Mistral | $0.0018 | $18.00 |
| Mistral Large 4 | Mistral | $0.0024 | $24.05 |
| Gemini 3.6 Flash | $0.0034 | $33.75 | |
| Gemini 3.7 Flash | $0.0034 | $33.75 | |
| Gemini 3.8 Flash | $0.0034 | $33.75 | |
| Nano Banana 2.1 | $0.0067 | $67.50 | |
| Mistral Medium 3.5 | Mistral | $0.0067 | $67.50 |
| Claude Sonnet 5.5 | Anthropic | $0.0090 | $90.00 |
| Gemini 4 Argon | $0.0090 | $90.00 | |
| GPT-6 Sol Pro | OpenAI | $0.0090 | $90.00 |
| GPT-6.1 Sol | OpenAI | $0.0090 | $90.00 |
| GPT-6.1 Sol Pro | OpenAI | $0.0090 | $90.00 |
| Command A | Cohere | $0.0100 | $100.00 |
| Command R+ (08-2024) | Cohere | $0.0100 | $100.00 |
| Claude Opus 5.5 | Anthropic | $0.0180 | $180.00 |
| Claude Opus 5 | Anthropic | $0.0225 | $225.00 |
| Claude Fable 5.1 | Anthropic | $0.0450 | $450.00 |
Frequently asked questions
How is the monthly cost calculated?
For each model, the tool multiplies your input tokens per request by the model’s input price per million tokens, does the same for output tokens, adds the two for the cost of one request, then multiplies by your requests per month.
Where do the prices come from, and how fresh are they?
Prices are read from OpenRouter’s public model catalogue and refreshed at most once a day. If that catalogue cannot be reached, the page falls back to a committed snapshot and labels how old it is. OpenRouter is a routing marketplace, so a model’s rate there can differ from the provider’s own list price. Check the vendor’s pricing page before relying on these figures in a contract.
What does the estimate leave out?
Each model uses one input price and one output price, so the tool does not model discounts such as prompt caching or batch processing, nor charges for tools, search, images or taxes. Providers generally bill reasoning or “thinking” tokens as output, so include them in your output-token figure.
How do I work out my tokens per request?
Paste a representative prompt into the Token Counter to get an input figure, and estimate a typical response length for the output figure. Remember that a system prompt and any conversation history you resend on every call count as input each time.