Checked November 2, 2026
Claude, Codex, and Cursor Token Prices
These are the public list rates the calculator multiplies by your token counts. Amounts are US dollars per million tokens. They are approximate, and the vendors can change them without notice.
Claude API
Short cache writes are the 5-minute price. Long cache writes are the 1-hour price. Cache reads on Fable 5.1 are $0.25, on Opus 5.5 are $0.20, and on Sonnet 5.5 are $0.10. Other Claude models use 0.1× the input price. Haiku 5.5 has a second card above 100,000 input-side tokens. US-only routing, where the model lists it, is a 1.1× multiplier you can turn on in the calculator.
| Model | Input | Cache read | Short cache write | Long cache write | Output | Notes |
|---|---|---|---|---|---|---|
| Claude Sonnet 5.5 | $2.00 / MTok | $0.10 / MTok | $2.50 / MTok | $4.00 / MTok | $10.00 / MTok | Cache reads are $0.10 / MTok (0.05× input) on Sonnet 5.5. |
| Claude Opus 5.5 | $4.00 / MTok | $0.20 / MTok | $5.00 / MTok | $8.00 / MTok | $20.00 / MTok | Cache reads are $0.20 / MTok (0.05× input). Fast: input $8.00 / MTok, output $40.00 / MTok. |
| Claude Haiku 5.5 | $0.10 / MTok | $0.01 / MTok | $0.125 / MTok | $0.20 / MTok | $0.50 / MTok | Prompts over 100,000 tokens are about 5× the standard token rates. Above 100,000 input-side tokens: input $0.50 / MTok, output $2.50 / MTok. |
| Claude Fable 5.1 | $10.00 / MTok | $0.25 / MTok | $12.50 / MTok | $20.00 / MTok | $50.00 / MTok | Cache reads are $0.25 / MTok (0.025× input). |
| Claude Mythos 5.1 | $10.00 / MTok | $0.25 / MTok | $12.50 / MTok | $20.00 / MTok | $50.00 / MTok | Limited availability. Same public token rates as Fable 5.1. |
| Claude Sonnet 5 | $2.00 / MTok | $0.20 / MTok | $2.50 / MTok | $4.00 / MTok | $10.00 / MTok | The $2 / $10 rate is the standard price. Cache reads stay at 0.1× input ($0.20). |
| Claude Opus 5 | $5.00 / MTok | $0.50 / MTok | $6.25 / MTok | $10.00 / MTok | $25.00 / MTok | Fast: input $10.00 / MTok, output $50.00 / MTok. |
| Claude Opus 4.8 | $5.00 / MTok | $0.50 / MTok | $6.25 / MTok | $10.00 / MTok | $25.00 / MTok | Fast: input $10.00 / MTok, output $50.00 / MTok. |
| Claude Opus 4.7 | $5.00 / MTok | $0.50 / MTok | $6.25 / MTok | $10.00 / MTok | $25.00 / MTok | |
| Claude Opus 4.6 | $5.00 / MTok | $0.50 / MTok | $6.25 / MTok | $10.00 / MTok | $25.00 / MTok | |
| Claude Sonnet 4.6 | $3.00 / MTok | $0.30 / MTok | $3.75 / MTok | $6.00 / MTok | $15.00 / MTok | |
| Claude Opus 4.5 | $5.00 / MTok | $0.50 / MTok | $6.25 / MTok | $10.00 / MTok | $25.00 / MTok | No US inference_geo multiplier on this card. That 1.1× factor starts with Claude 4.6. |
| Claude Sonnet 4.5 | $3.00 / MTok | $0.30 / MTok | $3.75 / MTok | $6.00 / MTok | $15.00 / MTok | |
| Claude Haiku 4.5 | $1.00 / MTok | $0.10 / MTok | $1.25 / MTok | $2.00 / MTok | $5.00 / MTok | |
| Claude Fable 5 | $10.00 / MTok | $1.00 / MTok | $12.50 / MTok | $20.00 / MTok | $50.00 / MTok | Cache reads are $1 / MTok, unlike Fable 5.1. |
| Claude Mythos 5 | $10.00 / MTok | $1.00 / MTok | $12.50 / MTok | $20.00 / MTok | $50.00 / MTok | Limited availability. Same public token rates as Fable 5. |
| Claude Opus 4.1 | $15.00 / MTok | $1.50 / MTok | $18.75 / MTok | $30.00 / MTok | $75.00 / MTok | Retired on the first-party card. Retired on the first-party API except some partner clouds. |
| Claude Opus 4 | $15.00 / MTok | $1.50 / MTok | $18.75 / MTok | $30.00 / MTok | $75.00 / MTok | Retired on the first-party card. Retired on the first-party API except some partner clouds. |
| Claude Sonnet 4 | $3.00 / MTok | $0.30 / MTok | $3.75 / MTok | $6.00 / MTok | $15.00 / MTok | Retired on the first-party card. Retired on the first-party API except some partner clouds. |
| Claude Haiku 3.5 | $0.80 / MTok | $0.08 / MTok | $1.00 / MTok | $1.60 / MTok | $4.00 / MTok | Retired on the first-party card. Retired on the first-party API except some partner clouds. |
OpenAI and Codex
Short cache writes are the 30-minute price. Long cache writes are the 12-hour price. Models with a dash in those columns do not publish a separate write price. Tokens entered there are counted at the input rate. Long-context rates, where published, apply above 272,000 input-side tokens to the whole request. GPT-5.3 Codex is the specialized API model. GPT-6.1 Sol, GPT-6 Luna, and GPT-6 Astra are the models called out for Codex with ChatGPT sign-in.
| Model | Input | Cache read | Short cache write | Long cache write | Output | Notes |
|---|---|---|---|---|---|---|
| GPT-6.1 Sol | $2.00 / MTok | $0.10 / MTok | $2.50 / MTok | $5.00 / MTok | $10.00 / MTok | Default coding model called out for ChatGPT Plus and Pro Codex sign-in. Above 272,000 input-side tokens: input $4.00 / MTok, output $15.00 / MTok. Fast: input $4.00 / MTok, output $20.00 / MTok. |
| GPT-6 Luna | $0.10 / MTok | $0.01 / MTok | $0.125 / MTok | $0.25 / MTok | $0.50 / MTok | Smaller Codex model used on Free and Go, and for higher-volume work on paid plans. Above 272,000 input-side tokens: input $0.20 / MTok, output $0.75 / MTok. Fast: input $0.20 / MTok, output $1.00 / MTok. |
| GPT-6 Sol | $2.00 / MTok | $0.20 / MTok | $2.50 / MTok | $5.00 / MTok | $10.00 / MTok | Above 272,000 input-side tokens: input $4.00 / MTok, output $15.00 / MTok. Fast: input $4.00 / MTok, output $20.00 / MTok. |
| GPT-6 Astra | $10.00 / MTok | $1.00 / MTok | $12.50 / MTok | $25.00 / MTok | $50.00 / MTok | Highest Codex tier on the public API card. Included usage on Pro is still not a token quota. Above 272,000 input-side tokens: input $20.00 / MTok, output $75.00 / MTok. Fast: input $20.00 / MTok, output $100.00 / MTok. |
| GPT-5.3 Codex | $1.75 / MTok | $0.175 / MTok | Not published | Not published | $14.00 / MTok | Specialized Codex API model. No separate cache-write price is published. Deprecated for ChatGPT sign-in; still listed on the API price card. Fast: input $3.50 / MTok, output $28.00 / MTok. |
| GPT-5.6 Sol | $4.00 / MTok | $0.40 / MTok | $5.00 / MTok | $10.00 / MTok | $20.00 / MTok | Promotional list price on the OpenAI card through at least November 21, 2026. Above 272,000 input-side tokens: input $8.00 / MTok, output $30.00 / MTok. Fast: input $8.00 / MTok, output $40.00 / MTok. |
| GPT-5.6 Terra | $2.00 / MTok | $0.20 / MTok | $2.50 / MTok | $5.00 / MTok | $12.00 / MTok | Above 272,000 input-side tokens: input $4.00 / MTok, output $18.00 / MTok. Fast: input $4.00 / MTok, output $24.00 / MTok. |
| GPT-5.6 Luna | $0.20 / MTok | $0.02 / MTok | $0.25 / MTok | $0.50 / MTok | $1.20 / MTok | Above 272,000 input-side tokens: input $0.40 / MTok, output $1.80 / MTok. Fast: input $0.40 / MTok, output $2.40 / MTok. |
| GPT-5.5 | $5.00 / MTok | $0.50 / MTok | Not published | Not published | $30.00 / MTok | Above 272,000 input-side tokens: input $10.00 / MTok, output $45.00 / MTok. Fast: input $12.50 / MTok, output $75.00 / MTok. |
| GPT-5.4 | $2.50 / MTok | $0.25 / MTok | Not published | Not published | $15.00 / MTok | Above 272,000 input-side tokens: input $5.00 / MTok, output $22.50 / MTok. Fast: input $5.00 / MTok, output $30.00 / MTok. |
| GPT-5.4 mini | $0.75 / MTok | $0.075 / MTok | Not published | Not published | $4.50 / MTok | No separate cache-write price is published. |
Cursor models
These rates are what Cursor publishes for Composer and Grok beyond the weekly included pool. Third-party models in Cursor use the Claude and OpenAI tables above. The Cursor Token Rate is $0.25 per million tokens on Teams and Enterprise third-party requests. It is not included unless you enable it, and it does not apply to Composer or Grok.
| Model | Input | Cache read | Short cache write | Long cache write | Output | Notes |
|---|---|---|---|---|---|---|
| Composer 2.5 | $0.50 / MTok | $0.20 / MTok | Not published | Not published | $2.50 / MTok | Cursor's own model. Rates are what Cursor publishes for usage beyond the weekly pool. No cache-write price. Fast: input $3.00 / MTok, output $15.00 / MTok. |
| Grok 4.7 | $2.00 / MTok | $0.50 / MTok | Not published | Not published | $6.00 / MTok | Included in the Cursor-model weekly pool. These rates apply past that pool. Above 256,000 input-side tokens: input $4.00 / MTok, output $12.00 / MTok. Fast: input $4.00 / MTok, output $12.00 / MTok. |
| Grok 4.6 | $2.00 / MTok | $0.50 / MTok | Not published | Not published | $6.00 / MTok | Fast: input $4.00 / MTok, output $12.00 / MTok. |
| Grok 4.5 | $2.00 / MTok | $0.50 / MTok | Not published | Not published | $6.00 / MTok | Fast output is $18 / MTok on Cursor's card. Fast: input $4.00 / MTok, output $18.00 / MTok. |
How to update this table
Edit lib/pricing.ts. Change the dollar amounts, set PRICING_UPDATED, and run npm test. The Claude Code cost calculator, the Codex and Cursor pages, and this table all read that file. Nothing is fetched at runtime.
- Anthropic model pricing
- Claude plans
- Claude Code on Pro, Max, and Ultra
- OpenAI API pricing
- ChatGPT plans
- Codex pricing
- Cursor models and pricing
- Cursor plans
Batch and flex discounts are documented by the vendors and are not applied here. Neither are taxes, credits, or per-request tool fees.