Groq AI inference capacity and cost calculator for agents
Enter the size of your content in characters and see how many tokens it is, whether it fits the context window of each model on Groq, and what it costs.
200,000characters
0Tokens
0Lines
0Pages
3Chars/token
How it works
- Character slider: ranges from 1,000 to about 3 million on a logarithmic scale. You can also type the exact number.
- Content type: English (4 characters per token), Spanish (3.5) or Code (3). Code is the default.
- Top indicators: approximate tokens, lines of code or words, pages (≈3,000 characters each) and characters per token.
- Reserved response: tokens you leave free for the reply (8,000 by default). The calculation adds input and reply against the context window.
- One card per model: shows whether it fits, the percentage of the window used, tokens left over or missing, input, output and total cost, and the maximum characters that fit.
- Output limit: if the reserved response exceeds the model's maximum (for example, 32,768 on Llama 3.3 70B), the card shows “Does not fit”.
- Estimate: real token counts depend on each model's tokenizer and can vary by ±15%.
Prices used
USD per million tokens (input / output):
- Llama 3.1 8B Instant: $0.05 / $0.08
- Llama 3.3 70B Versatile: $0.59 / $0.79
- GPT-OSS 20B: $0.075 / $0.30
- GPT-OSS 120B: $0.15 / $0.60
- MiniMax M2.7: Enterprise (price through sales contact)
The values come from public Groq listings (October 2026) and the sources do not fully agree: some say Llama and MiniMax are Enterprise-only, others show the earlier list prices. Confirm them in the Groq console.
Part of Refactorii