AI API Price Index
What 15 language models from 5 providers cost through their APIs, in US dollars per 1 million tokens. Every price comes from the provider's own pricing page, is checked daily, and goes live only after an editor reviews the change.
Estimate your monthly bill on every model: AI API cost calculator
Prices per 1M tokens and cost of 1,000 typical tasks
Standard pay-as-you-go prices, without batch or priority tiers. Task costs use the same token counts for every model: chat reply — 1,500 tokens in, 400 out; answer over documents (RAG) — 8,000 tokens in, 500 out; agent step with 90% of context cached — 30,000 tokens in, 1,000 out.
Anthropic
| Model | Input | Cached input | Output | Context | Chat ×1K | RAG ×1K | Agent ×1K |
|---|---|---|---|---|---|---|---|
Claude Haiku 5.5 Prompts up to 100,000 tokens; higher prices for longer prompts | $0.10 | $0.01 | $0.50 | — | $0.35 | $1.05 | $1.07 |
Claude Sonnet 5.5 | $2.00 | $0.10 | $10.00 | — | $7.00 | $21.00 | $18.70 |
Claude Sonnet 5 | $2.00 | $0.20 | $10.00 | — | $7.00 | $21.00 | $21.40 |
Claude Opus 5.5 | $4.00 | $0.20 | $20.00 | — | $14.00 | $42.00 | $37.40 |
Claude Opus 5 | $5.00 | $0.50 | $25.00 | — | $17.50 | $52.50 | $53.50 |
Claude Fable 5 | $10.00 | $1.00 | $50.00 | — | $35.00 | $105.00 | $107.00 |
Claude Fable 5.1 | $10.00 | $0.25 | $50.00 | — | $35.00 | $105.00 | $86.75 |
DeepSeek
| Model | Input | Cached input | Output | Context | Chat ×1K | RAG ×1K | Agent ×1K |
|---|---|---|---|---|---|---|---|
deepseek-flash Peak-hour rates (standard); off-peak rates are half: input $0.15, output $0.6, cache hit $0.003. Peak hours 01:00-04:00 and 06:00-10:00 UTC, Mon-Fri. | $0.30 | $0.006 | $1.20 | 1M | $0.93 | $3.00 | $2.26 |
| Model | Input | Cached input | Output | Context | Chat ×1K | RAG ×1K | Agent ×1K |
|---|---|---|---|---|---|---|---|
gemini-3.5-flash-lite | $0.30 | $0.03 | $2.50 | — | $1.45 | $3.65 | $4.21 |
gemini-3.6-flash Price through Dec 31, 2026; rises to $1.50 input / $7.50 output from Jan 1, 2027. | $0.75 | $0.075 | $3.75 | — | $2.63 | $7.88 | $8.02 |
OpenAI
| Model | Input | Cached input | Output | Context | Chat ×1K | RAG ×1K | Agent ×1K |
|---|---|---|---|---|---|---|---|
gpt-6-luna | $0.10 | $0.01 | $0.50 | — | $0.35 | $1.05 | $1.07 |
gpt-6.1-terra | $1.00 | $0.05 | $5.00 | — | $3.50 | $10.50 | $9.35 |
gpt-6.1-sol | $2.00 | $0.10 | $10.00 | — | $7.00 | $21.00 | $18.70 |
gpt-6-astra | $10.00 | $1.00 | $50.00 | — | $35.00 | $105.00 | $107.00 |
xAI
| Model | Input | Cached input | Output | Context | Chat ×1K | RAG ×1K | Agent ×1K |
|---|---|---|---|---|---|---|---|
grok-4.7 | $2.00 | — | $6.00 | 500K | $5.40 | $19.00 | $66.00 |
Recent price changes
- OpenAI gpt-6.1-sol: price details updated per 1M tokens
- OpenAI gpt-6.1-terra: price details updated per 1M tokens
- OpenAI gpt-6-luna: price details updated per 1M tokens
- OpenAI gpt-6-astra: price details updated per 1M tokens
- Anthropic Claude Fable 5.1: price details updated per 1M tokens
- Anthropic Claude Opus 5.5: price details updated per 1M tokens
- Anthropic Claude Sonnet 5.5: price details updated per 1M tokens
- Anthropic Claude Haiku 5.5: price details updated per 1M tokens
- Anthropic Claude Fable 5: price details updated per 1M tokens
- Anthropic Claude Opus 5: price details updated per 1M tokens
- Anthropic Claude Sonnet 5: price details updated per 1M tokens
- OpenAI gpt-6-luna added: input $0.10, output $0.50 per 1M tokens
- Anthropic Claude Sonnet 5 added: input $2.00, output $10.00 per 1M tokens
- OpenAI gpt-6.1-terra added: input $1.00, output $5.00 per 1M tokens
- xAI grok-4.7 added: input $2.00, output $6.00 per 1M tokens
- OpenAI gpt-6.1-sol added: input $2.00, output $10.00 per 1M tokens
- Anthropic Claude Fable 5 added: input $10.00, output $50.00 per 1M tokens
- Anthropic Claude Fable 5.1 added: input $10.00, output $50.00 per 1M tokens
- Anthropic Claude Haiku 5.5 added: input $0.10, output $0.50 per 1M tokens
- Anthropic Claude Opus 5 added: input $5.00, output $25.00 per 1M tokens
How the index is built
- Sources. Only official pricing pages of the providers. Third-party aggregators are used to cross-check numbers, never as a source.
- Checks. A crawler reads each pricing page daily. Every price must match a quoted line on that page; anything new or different waits for an editor before it is published.
- Which price. Standard pay-as-you-go price per 1 million tokens in US dollars. Where the price depends on prompt length, we show the lowest tier and say so under the model name. Cached input means reading from the prompt cache.
- Tokens are not equal. Each provider splits text into tokens differently, so the same text can be a different number of tokens on different models. Compare task costs as a guide to price levels, not as an exact bill.
- Not included. Taxes, volume discounts, free tiers, batch and priority pricing, tool and search fees, and image, audio and embedding models.
Use the data
The index is free to reuse under CC BY 4.0: copy, chart and republish it, including commercially, with a link back. Download JSON or CSV, or follow price changes by RSS.
Attribution to use with a chart or table:
Source: <a href="https://modelalmanac.com/ai-api-prices/">Model Almanac AI API Price Index</a>, CC BY 4.0