AI model prices,
side by side.
Public API list prices for 17 current text models, each linked to the provider’s own pricing page.
Three prices matter for most apps.
Input is what you send. Cached input is the lower rate for a repeated prompt prefix the provider has stored. Output is what the model writes, including billed reasoning tokens. Prices are in US dollars per one million tokens. Among the listed models, GPT-6 Luna has the lowest combined input and output list price, but the right model is the cheapest one that passes your own quality test.
OpenAI
OpenAI API pricing ↗| Model | Tier | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
GPT-6 Astragpt-6-astra | Frontier | $10.00 | $1.00 | $50.00 | Over 272,000 input tokens: $20.00 in / $75.00 out. |
GPT-6.1 Solgpt-6.1-sol | Balanced | $2.00 | $0.10 | $10.00 | Over 272,000 input tokens: $4.00 in / $15.00 out. |
GPT-6.1 Terragpt-6.1-terra | Balanced | $1.00 | $0.05 | $5.00 | Over 272,000 input tokens: $2.00 in / $7.50 out. |
GPT-6 Lunagpt-6-luna | Fast & low-cost | $0.10 | $0.01 | $0.50 | Over 272,000 input tokens: $0.20 in / $0.75 out. |
GPT-5.4 minigpt-5.4-mini | Fast & low-cost | $0.75 | $0.075 | $4.50 | — |
Batch API: Batch and Flex rates are listed at about half the Standard rate.
Anthropic
Claude API pricing ↗| Model | Tier | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
Claude Fable 5.2claude-fable-5-2 | Frontier | $8.00 | $0.20 | $40.00 | — |
Claude Opus 5.5claude-opus-5-5 | Frontier | $4.00 | $0.20 | $20.00 | — |
Claude Sonnet 5.5claude-sonnet-5-5 | Balanced | $2.00 | $0.10 | $10.00 | — |
Claude Haiku 5.5claude-haiku-5-5 | Fast & low-cost | $0.10 | $0.01 | $0.50 | Over 100,000 input tokens: $0.50 in / $2.50 out. |
Batch API: The Batch API gives a 50% discount on input and output tokens.
| Model | Tier | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
Gemini 4 Carbongemini-4-carbon | Frontier | $2.00 | $0.10 | $10.00 | From December 1, 2026: $4.00 in / $20.00 out. |
Gemini 3.1 Pro (preview)gemini-3.1-pro-preview | Frontier | $2.00 | $0.20 | $12.00 | Over 200,000 input tokens: $4.00 in / $18.00 out.Context caching also has an hourly storage charge. |
Gemini 3.8 Flashgemini-3.8-flash | Balanced | $0.75 | $0.075 | $3.75 | From January 1, 2027: $1.50 in / $7.50 out.Context caching also has an hourly storage charge. |
Gemini 3.5 Flash-Litegemini-3.5-flash-lite | Fast & low-cost | $0.30 | $0.03 | $2.50 | Context caching also has an hourly storage charge. |
Gemini 3.1 Flash-Litegemini-3.1-flash-lite | Fast & low-cost | $0.25 | $0.025 | $1.50 | Text, image and video input price. Audio input costs more. |
Batch API: The Batch API is listed at a 50% cost reduction.
DeepSeek
DeepSeek API pricing ↗| Model | Tier | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
DeepSeek V4 Prodeepseek-v4-pro | Balanced | $1.32 | $0.044 | $3.96 | — |
DeepSeek Flashdeepseek-flash | Fast & low-cost | $0.30 | $0.006 | $1.20 | — |
Off-peak hours: Off-peak rates are half the peak rates. Peak is 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday.
Mistral
Mistral pricing ↗| Model | Tier | Input | Cached input | Output | Notes |
|---|---|---|---|---|---|
Mistral Largemistral-large | Balanced | $0.50 | — | $1.50 | The public page lists no separate cached-input rate for this model. |
Batch processing: Batch processing is listed at a 50% reduction.
USD per 1 million tokens, standard tier, checked November 12, 2026. Selected current text models, not every model or tier a provider offers. DeepSeek prices are peak-hour rates. Confirm prices on the provider’s page before you commit a budget.
Compare on your own task.
A cheaper model that needs two retries or a human fix can cost more than a stronger one. Run the same small set of real tasks through two or three candidates. Record accuracy, output length and latency alongside the token bill.
Read: a checklist for choosing an LLM API →Prices and models move.
Providers publish price increases, retire model names and route old names to new models. Pin a model version in production, watch the provider’s deprecation notices and re-check this table before renewing a budget.
Read: handling deprecations and price changes →