LLM API cost calculator.
Describe a workload and see what it costs per month on 22 models. Prices come from my LLM API pricing tracker, last checked 12 November 2026.
That is 200M input tokens and 50M output tokens a month.
| Model | Input | Cached input | Output | Per month |
|---|---|---|---|---|
| Ministral 3 8B | $30.00 | $0.00 | $7.50 | $37.50 |
| Google gemini-2.5-flash-lite | $20.00 | $0.00 | $20.00 | $40.00 |
| OpenAI gpt-6-luna | $20.00 | $0.00 | $25.00 | $45.00 |
| Claude Haiku 5.5prompts up to 100,000 tokens | $20.00 | $0.00 | $25.00 | $45.00 |
| Mistral Small 4 | $30.00 | $0.00 | $30.00 | $60.00 |
| DeepSeek deepseek-flashpeak price shown; input $0.30 peak, $0.15 off-peak, output $1.20 peak, $0.60 off-peak | $60.00 | $0.00 | $60.00 | $120 |
| Google gemini-3.1-flash-lite | $50.00 | $0.00 | $75.00 | $125 |
| Google gemini-3.5-flash-lite | $60.00 | $0.00 | $125 | $185 |
| Mistral Large 4 | $140 | $0.00 | $105 | $245 |
| Google gemini-3.8-flash | $150 | $0.00 | $188 | $338 |
| xAI grok-4.3 | $250 | $0.00 | $125 | $375 |
| OpenAI gpt-6.1-terra | $200 | $0.00 | $250 | $450 |
| DeepSeek deepseek-v4-propeak price shown; input $1.32 peak, $0.66 off-peak, output $3.96 peak, $1.98 off-peak | $264 | $0.00 | $198 | $462 |
| Google gemini-4-hydrogen | $300 | $0.00 | $375 | $675 |
| xAI grok-4.7 | $400 | $0.00 | $300 | $700 |
| OpenAI gpt-6.1-sol | $400 | $0.00 | $500 | $900 |
| Claude Sonnet 5.5 | $400 | $0.00 | $500 | $900 |
| Google gemini-4-carbon | $400 | $0.00 | $500 | $900 |
| Google gemini-4-argon | $400 | $0.00 | $500 | $900 |
| Claude Opus 5.5 | $800 | $0.00 | $1,000 | $1,800 |
| Claude Fable 5.2 | $1,600 | $0.00 | $2,000 | $3,600 |
| OpenAI gpt-6-astra | $2,000 | $0.00 | $2,500 | $4,500 |
What the numbers include
- Standard-tier list prices in US dollars per million tokens, as each vendor publishes them.
- Cached input is priced at the vendor's cache-read rate. Cache-write surcharges and cache storage fees are not included.
- Long-context surcharges, batch discounts, fast or priority tiers and scheduled price changes are not included. The tracker lists them.
- Where a vendor has peak and off-peak prices, the peak price is used and the row says so.
- Output tokens include thinking tokens on models that bill for them.
This is arithmetic on list prices, not a quote. Your invoice also depends on retries, tool calls and how much of your prompt really hits the cache. For ways to bring the number down, see cutting LLM API costs with caching and batching and prompt caching explained.
Spot a wrong price? Email hello@ajayk.sh.