AI model pricing, compared.
Input and output cost per million tokens for the flagship text models, sourced from each provider's own pricing page. Pick a model for the full breakdown and a head-to-head comparison.
A dated record of every price move we verify, including the introductory rates that already have a published end date. The prices above are a snapshot. This is the trend.
See the change logOne table, every provider.
| Model | Provider | Input / 1M | Output / 1M |
|---|---|---|---|
| Claude Sonnet 5 | Anthropic | $2.00 | $10.00 |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 |
| Claude Haiku 4.5 | Anthropic | $1.00 | $5.00 |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 |
| GPT-5 | OpenAI | $1.25 | $10.00 |
| GPT-5.6 Sol | OpenAI | $4.00 | $20.00 |
| Gemini 3.1 Pro | $2.00 | $12.00 | |
| Gemini 3.8 FlashIntroductory rate through 31 Dec 2026 | $0.75 | $3.75 | |
| DeepSeek V4 ProOff-peak rate. Peak (01:00-04:00 and 06:00-10:00 UTC, Mon-Fri) is 2x | DeepSeek | $0.66 | $1.98 |
| Grok 4.6Rate doubles on prompts of 200k tokens or more | xAI | $2.00 | $6.00 |
| Llama 3.3 70BGroq no longer publishes a per-token price for this model (contact sales), so this is the Together rate | Together | $1.04 | $1.04 |
| Mistral Large | Mistral | $0.50 | $1.50 |
Common questions.
How often is this pricing checked?
The date at the top of each model's page is when it was last checked against the provider's own pricing page. This index shows 2026-09-05. AI provider pricing changes without much notice, always cross-check the linked source before making a budget commitment based on these numbers.
Why do you show price per million tokens instead of per request?
Per-million-token is the unit every provider actually bills in, and it is the only unit that lets you compare providers honestly since request cost depends entirely on how many tokens your specific task uses. Each model page also shows a sample-request cost for a typical 1,200-in/400-out exchange so the numbers are concrete.