Workloads
App cost estimator
Start from a workload that looks like yours. The assumptions are written out, and every number can be changed. The figure at the top is the cheapest model in the list for that usage.
What will a support bot cost?
Gemini 2.5 Flash-Lite is the cheapest of these, at $2.81 a month.
A month is 30 days. These assumptions are a starting point. Change them.
Short replies, many times a day, with a chunk of the prompt reused from a cache.
650 chats a day. 800 input tokens, of which 400 are cached. 250 output tokens. Change any of these.
| Model | One request | A month |
|---|---|---|
| Gemini 2.5 Flash-Lite | $0.0001 | $2.81 |
| GPT-6 Luna | $0.0002 | $3.30 |
| Claude Haiku 5.5 | $0.0002 | $3.30 |
| DeepSeek Flash | $0.0002 | $4.12 |
| Mistral Small 4 | $0.0002 | $4.21 |
| DeepSeek V4 Pro | $0.0008 | $14.97 |
| Grok 4.3 | $0.0012 | $23.50 |
| Gemini 3.8 Flash | $0.0013 | $24.72 |
| Mistral Large 4 | $0.0016 | $32.08 |
| Grok 4.7 | $0.0025 | $48.75 |
| GPT-6.1 Sol | $0.0033 | $65.13 |
| Claude Sonnet 5.5 | $0.0033 | $65.13 |
| Gemini 4 Argon | $0.0033 | $65.13 |
| Claude Fable 5.1 | $0.02 | $323.70 |
| GPT-6 Astra | $0.02 | $329.55 |