How much does an AI API cost in 2026?
AI API prices in 2026 run from $0.10 to $10 per million input tokens across OpenAI, Claude, Gemini and DeepSeek. Output tokens cost 3 to 8 times more than input. GPT-6 Luna and Claude Haiku 5.5 are the cheapest at $0.10 input and $0.50 output, a fit for developers who run high volumes. Updated October 2026.
- Cheapest for high volume: GPT-6 Luna and Claude Haiku 5.5 cost $0.10 input and $0.50 output. DeepSeek V4.1 Flash costs $0.15 and $0.60 off-peak, and $0.30 and $1.20 at peak.
- Everyday work: GPT-6.1 Sol and Claude Sonnet 5.5 cost $2 input and $10 output. Gemini 4 Argon also costs $2 and $10 until 30 November 2026, then $4 and $20.
- Top models: GPT-6 Astra and Claude Fable 5.1 cost $10 input and $50 output. Claude Opus 5.5 costs $4 and $20.
- Repeated prompts: cached input costs from $0.003 per million tokens on DeepSeek V4.1 Flash off-peak, and $0.01 on GPT-6 Luna and Claude Haiku 5.5.
- Work that can wait: OpenAI, Claude and Gemini take 50% off in batch mode. DeepSeek halves the price at off-peak hours instead.
All prices are US dollars per 1 million tokens and were current on 23 October 2026. A token is a small piece of text such as a word, a number or a punctuation mark.
Who is this AI API pricing guide for?
This guide is for developers, product leads and founders who budget a monthly bill for an AI model API. It sets the current OpenAI, Claude, Gemini and DeepSeek models side by side, with the extra charges that change a bill: cached input, batch mode, long prompts and search tools.
It covers the API, which bills per token. The chat apps bill a flat monthly fee, and those prices sit on separate guides: Claude subscription prices, ChatGPT plan costs, Gemini plans for individuals and Grok plans.
How much does each AI API cost per million tokens?
OpenAI, Claude, Gemini and DeepSeek charge $0.10 to $10 per million input tokens and $0.50 to $50 per million output tokens on the models below.
| Model | Input | Cached input | Output | Context | Batch (in / out) |
|---|---|---|---|---|---|
| GPT-6 Astra | $10 | $1.00 | $50 | 1.05M | $5 / $25 |
| GPT-6.1 Sol | $2 | $0.10 | $10 | 1.05M | $1 / $5 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 | 1.05M | $0.05 / $0.25 |
| Claude Fable 5.1 | $10 | $0.25 | $50 | 1M | $5 / $25 |
| Claude Opus 5.5 | $4 | $0.20 | $20 | 1M | $2 / $10 |
| Claude Sonnet 5.5 | $2 | $0.10 | $10 | 1M | $1 / $5 |
| Claude Haiku 5.5, up to 100K | $0.10 | $0.01 | $0.50 | 1M | $0.05 / $0.25 |
| Claude Haiku 5.5, over 100K | $0.50 | $0.05 | $2.50 | 1M | $0.25 / $1.25 |
| Gemini 4 Argon, to 30 Nov | $2 | $0.10 | $10 | 1M | $1 / $5 |
| Gemini 4 Argon, from 1 Dec | $4 | $0.20 | $20 | 1M | $2 / $10 |
| Gemini 3.8 Flash, to 31 Dec | $0.75 | $0.075 | $3.75 | 1M | $0.375 / $1.875 |
| Gemini 3.8 Flash, from 1 Jan | $1.50 | $0.15 | $7.50 | 1M | $0.75 / $3.75 |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 | 1M | $0.15 / $1.25 |
| DeepSeek V4.1 Flash, off-peak | $0.15 | $0.003 | $0.60 | 1M | No batch rate |
| DeepSeek V4.1 Flash, peak | $0.30 | $0.006 | $1.20 | 1M | No batch rate |
| DeepSeek V4 Pro, off-peak | $0.66 | $0.022 | $1.98 | 1M | No batch rate |
| DeepSeek V4 Pro, peak | $1.32 | $0.044 | $3.96 | 1M | No batch rate |
OpenAI prices are for prompts up to 272K input tokens, and the OpenAI context window is 1,050,000 tokens. The cached input column is the price of a cache hit. The two Claude Haiku 5.5 rows differ by prompt length, which sets the price. Gemini 4 Argon changes price on 1 December 2026 and Gemini 3.8 Flash on 1 January 2027, and DeepSeek peak prices are double the off-peak prices.
What does the OpenAI API cost?
GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens, GPT-6.1 Sol costs $2 and $10, and GPT-6 Luna costs $0.10 and $0.50.
OpenAI API
- Prompts over 272K tokens cost 2x input and 1.5x output
- Fast mode costs 2x and regional processing adds 10%
All three models have a 1,050,000-token context window and a 128,000-token output limit. Cached input costs $1.00, $0.10 and $0.01 per million tokens on Astra, Sol and Luna.
Batch and Flex processing cost 50% of the standard rate, which makes GPT-6.1 Sol $1 input and $5 output. Fast mode costs 2 times the standard rate, and Ultrafast mode on GPT-6.1 Sol costs 6 times. Writing a prompt into the cache costs 1.25 times the input rate for a 30-minute cache.
These three models have no free tier. OpenAI sets rate limits by usage tier (Build, Launch and Grow), and a tier rises as spend grows. On GPT-6 Astra the Build tier allows 5,000 requests and 1,000,000 tokens a minute.
What does the Claude API cost?
Claude API prices run from $0.10 to $10 per million input tokens and $0.50 to $50 per million output tokens, from Haiku 5.5 up to Fable 5.1.
Claude API (Anthropic)
- Haiku 5.5 prompts over 100K tokens cost $0.50 input
- US-only inference adds 10% to every token price
Claude Fable 5.1 costs $10 and $50, Claude Opus 5.5 costs $4 and $20, Claude Sonnet 5.5 costs $2 and $10, and Claude Haiku 5.5 costs $0.10 and $0.50. Every current Claude model has a 1M-token context window. Fable 5.1, Opus 5.5 and Sonnet 5.5 keep one price across the whole window, so a 900,000-token prompt costs the same per token as a 9,000-token prompt. Haiku 5.5 charges $0.50 input and $2.50 output when the prompt is over 100,000 tokens.
Batch mode takes 50% off, which makes Sonnet 5.5 $1 input and $5 output. A cache hit costs 0.25, 0.20, 0.10 and 0.01 dollars per million tokens on Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Opus 5.5 also has a fast mode at $8 input and $40 output, which is not available in batch mode.
New accounts get a small amount of free credits to test the API. Claude models from version 4.7 on use a newer tokenizer that produces about 30% more tokens for the same text than earlier Claude models, so count tokens on the model you plan to buy.
What does the Gemini API cost?
Gemini API prices run from $0.30 input and $2.50 output on 3.5 Flash-Lite to $2 and $10 on 4 Argon, which doubles on 1 December 2026.
Gemini API (Google)
- Argon prices double on 1 December 2026
- Gemini 3.8 Flash prices double on 1 January 2027
Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens until 30 November 2026, then $4 and $20. Gemini 3.8 Flash costs $0.75 input and $3.75 output through 31 December 2026, then $1.50 and $7.50 from 1 January 2027. All three models have a 1,048,576-token context window.
Gemini 3.8 Flash and Gemini 3.5 Flash-Lite have a free tier at no charge, and on the free tier Google uses the content to improve its products. Gemini 4 Argon has no free tier. The paid tier does not use content to improve Google products and adds higher rate limits, context caching and a Batch API with a 50% cost reduction.
Caching on Gemini 3.8 Flash costs $0.075 per million tokens through 31 December 2026 plus $0.50 per million tokens per hour of storage. Gemini 4 Argon uses implicit caching only, at $0.10 per million tokens until 30 November 2026.
What does the DeepSeek API cost?
DeepSeek V4.1 Flash costs $0.15 per million input tokens and $0.60 per million output tokens off-peak, and DeepSeek V4 Pro costs $0.66 and $1.98 off-peak.
DeepSeek API
- Off-peak hours cost half the peak rate
- Peak runs weekdays 01:00 to 04:00 and 06:00 to 10:00 UTC
Peak prices are double. Peak hours run Monday to Friday from 01:00 to 04:00 and 06:00 to 10:00 UTC, excluding Chinese public holidays, and every other hour is off-peak. At peak, V4.1 Flash costs $0.30 and $1.20 and V4 Pro costs $1.32 and $3.96.
Both models have a 1M-token context window and a 384K output limit. Context caching is on by default, and a cache hit costs $0.003 per million tokens on V4.1 Flash and $0.022 on V4 Pro off-peak. There is no batch rate.
Search and Fetch are separate APIs at $0.75 per 1,000 successful requests each. Billing for the server-side search and fetch calls made through DeepSeek Harness, Claude Code or Codex starts on 1 November 2026. Charges come out of a topped-up balance, and any granted balance is used first. The concurrency limit is 2,500 on V4.1 Flash and 500 on V4 Pro.
What do cached input and batch calls cost?
Cached input costs 0.02 to 0.1 times the normal input price, and batch mode costs half the standard price on OpenAI, Claude and Gemini.
| Provider | Cached input (cache hit) | Cache write | Batch |
|---|---|---|---|
| OpenAI | 5% of input on GPT-6.1 Sol, 10% on GPT-6 Astra and GPT-6 Luna | 1.25 times input for 30 minutes, 2.5 times for 12 hours on GPT-6.1 Sol | 50% off (Batch and Flex) |
| Claude | 0.025 times input on Fable 5.1, 0.05 times on Opus 5.5 and Sonnet 5.5, 0.1 times on Haiku 5.5 | 1.25 times input for 5 minutes, 2 times for 1 hour | 50% off |
| Gemini | Argon: $0.10 to 30 Nov 2026, then $0.20. 3.8 Flash: $0.075, then $0.15. 3.5 Flash-Lite: $0.03 | Storage per 1M tokens per hour: 3.8 Flash $0.50 to 31 Dec 2026, then $1.00; 3.5 Flash-Lite $1.00 | 50% off on the paid tier |
| DeepSeek | $0.003 (V4.1 Flash) and $0.022 (V4 Pro) off-peak, double at peak | No separate cache charge | No batch rate; off-peak is half of peak |
Cached input pays off on repeated prompts, such as a long system prompt sent with every support reply. On Claude, a 5-minute cache write pays for itself after one cache read, and a 1-hour write after two.
What costs extra on top of the token price?
OpenAI charges 2 times the input price above 272K tokens, and every provider adds charges for search tools, fast modes or data residency.
- Long prompts. OpenAI charges 2 times the input price and 1.5 times the output price for the whole request when the prompt is over 272K tokens, so GPT-6 Astra goes from $10 and $50 to $20 and $75. Claude Fable 5.1, Opus 5.5 and Sonnet 5.5 keep one price to 1M tokens, and Claude Haiku 5.5 steps up above 100K tokens.
- Search tools. Web search costs $10 per 1,000 calls on OpenAI and $10 per 1,000 searches on Claude. Gemini grounding with Google Search gives 5,000 free requests a month across Gemini 3 and newer models, then $14 per 1,000. DeepSeek Search costs $0.75 per 1,000 successful requests.
- Fast modes. OpenAI Fast mode is 2 times the standard rate, and Claude Opus 5.5 fast mode is $8 input and $40 output.
- Data residency. OpenAI regional processing adds 10%, and Claude US-only inference adds 10% on every token price.
- Code and file tools. OpenAI file search tool calls cost $2.50 per 1,000, and Claude code execution without web search or fetch gives 1,550 free hours a month per organization, then $0.05 an hour per container.
- Rate limits. Limits cap how much you can send a minute and do not change the price. They rise with spend on OpenAI and with the usage tier on Claude, and DeepSeek sets a concurrency limit per model.
What does a month of real use cost?
A support bot that answers 1,000 customer messages a day costs $10.50 a month on GPT-6 Luna and $210 a month on GPT-6.1 Sol, using the same traffic.
The example sends 2,000 input tokens (the system prompt, the conversation and the question) and gets back 300 output tokens for each reply. That is 30,000 replies, 60 million input tokens and 9 million output tokens a month, at standard prices with no caching.
| Model | Input cost | Output cost | Monthly total |
|---|---|---|---|
| GPT-6 Luna | 60M x $0.10 = $6.00 | 9M x $0.50 = $4.50 | $10.50 |
| Claude Haiku 5.5 (prompts up to 100K tokens) | 60M x $0.10 = $6.00 | 9M x $0.50 = $4.50 | $10.50 |
| DeepSeek V4.1 Flash, off-peak | 60M x $0.15 = $9.00 | 9M x $0.60 = $5.40 | $14.40 |
| DeepSeek V4.1 Flash, peak | 60M x $0.30 = $18.00 | 9M x $1.20 = $10.80 | $28.80 |
| Gemini 3.5 Flash-Lite | 60M x $0.30 = $18.00 | 9M x $2.50 = $22.50 | $40.50 |
| Gemini 3.8 Flash to 31 Dec 2026 | 60M x $0.75 = $45.00 | 9M x $3.75 = $33.75 | $78.75 |
| Gemini 3.8 Flash from 1 Jan 2027 | 60M x $1.50 = $90.00 | 9M x $7.50 = $67.50 | $157.50 |
| GPT-6.1 Sol | 60M x $2 = $120.00 | 9M x $10 = $90.00 | $210.00 |
| Claude Sonnet 5.5 | 60M x $2 = $120.00 | 9M x $10 = $90.00 | $210.00 |
| Claude Opus 5.5 | 60M x $4 = $240.00 | 9M x $20 = $180.00 | $420.00 |
| GPT-6 Astra | 60M x $10 = $600.00 | 9M x $50 = $450.00 | $1,050.00 |
| Claude Fable 5.1 | 60M x $10 = $600.00 | 9M x $50 = $450.00 | $1,050.00 |
Batch mode halves these totals when the replies can wait: the same 30,000 replies on Claude Sonnet 5.5 in batch mode are 60 million tokens at $1 plus 9 million tokens at $5, which is $105 a month. Gemini 4 Argon matches GPT-6.1 Sol at $210 a month until 30 November 2026 and costs $420 from 1 December 2026.
Which AI API fits your situation?
Prices run from $0.10 to $10 per million input tokens, so the job decides the model.
An AI API for a startup on a tight budget
GPT-6 Luna and Claude Haiku 5.5 cost $0.10 input and $0.50 output, and DeepSeek V4.1 Flash costs $0.15 and $0.60 off-peak. Gemini 3.8 Flash and Gemini 3.5 Flash-Lite also have a free tier for testing.
An AI API for high-volume support, tagging or extraction
Small models at $0.10 to $0.30 input handle repeated, short jobs. Cached input makes a long system prompt almost free: $0.01 per million tokens on GPT-6 Luna and Claude Haiku 5.5, and $0.003 on DeepSeek V4.1 Flash off-peak.
An AI API for coding agents and hard reasoning
GPT-6.1 Sol and Claude Sonnet 5.5 cost $2 input and $10 output. Claude Opus 5.5 costs $4 and $20, and GPT-6 Astra and Claude Fable 5.1 cost $10 and $50 for the hardest work.
An AI API for long documents
Every model here reads at least 1 million tokens. Claude Fable 5.1, Opus 5.5 and Sonnet 5.5 keep one price to 1M tokens, while OpenAI charges 2 times the input price above 272K tokens.
An AI API for overnight jobs and bulk processing
Batch mode takes 50% off on OpenAI, Claude and Gemini. A bulk summary job on Claude Sonnet 5.5 costs $1 input and $5 output in batch mode instead of $2 and $10.
An AI API for a team with data residency needs
OpenAI regional processing adds 10% and Claude US-only inference adds 10%. Plan for that extra 10% in the budget before you pick a region.
How do you estimate your own AI API bill?
Four steps turn the $0.10 to $10 price range into a monthly number for your own traffic.
- Count the tokens in one typical request and one typical answer, using the token counter of the model you plan to buy.
- Multiply the requests a month by the input tokens, then by the input price, and do the same for output.
- Subtract the saving from cached input and batch mode if your traffic allows them.
- Add the search, fast mode or residency charges, then test two models on real traffic for a week before you commit.
Output tokens drive the bill when answers are long, and input tokens drive it when prompts are long.
Best for your situation
Guides & comparisons
Frequently asked questions
How much does the Claude API cost?
The Claude API costs $0.10 to $10 per million input tokens and $0.50 to $50 per million output tokens. Claude Haiku 5.5 costs $0.10 and $0.50, Claude Sonnet 5.5 costs $2 and $10, Claude Opus 5.5 costs $4 and $20, and Claude Fable 5.1 costs $10 and $50. Batch mode takes 50% off.
How much does the OpenAI API cost?
The OpenAI API costs $0.10 to $10 per million input tokens and $0.50 to $50 per million output tokens. GPT-6 Luna costs $0.10 and $0.50, GPT-6.1 Sol costs $2 and $10, and GPT-6 Astra costs $10 and $50. Prompts over 272K input tokens cost 2 times the input price and 1.5 times the output price.
Is the OpenAI API free?
No. The GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna models have no free tier, and usage bills per token. Batch and Flex processing cost 50% less.
Is the Claude API free?
No. Claude bills per token, and new accounts get a small amount of free credits to test the API. The cheapest model, Claude Haiku 5.5, costs $0.10 per million input tokens.
Is the DeepSeek API free?
No. DeepSeek bills per token from a topped-up balance, and a granted balance is used first when an account has one. DeepSeek V4.1 Flash costs $0.15 per million input tokens and $0.60 per million output tokens off-peak.
Does the Gemini API have a free tier?
Yes for two models. Gemini 3.8 Flash and Gemini 3.5 Flash-Lite are free of charge on the free tier, and Google uses that content to improve its products. Gemini 4 Argon is not available on the free tier.
What is the cheapest Claude API model?
Claude Haiku 5.5 is the cheapest Claude model at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Above 100,000 tokens it costs $0.50 and $2.50.
Is the Claude API cheaper than the ChatGPT API?
The prices match in three pairs. Claude Haiku 5.5 and GPT-6 Luna cost $0.10 and $0.50, Claude Sonnet 5.5 and GPT-6.1 Sol cost $2 and $10, and Claude Fable 5.1 and GPT-6 Astra cost $10 and $50. Above 272K tokens OpenAI charges more, while Claude Sonnet 5.5 and Opus 5.5 keep one price.
Is the DeepSeek API cheaper than OpenAI?
DeepSeek V4 Pro costs $0.66 input and $1.98 output off-peak, which is less than GPT-6.1 Sol at $2 and $10. DeepSeek V4.1 Flash costs $0.15 and $0.60 off-peak, and GPT-6 Luna costs $0.10 and $0.50. DeepSeek peak prices are double its off-peak prices.
Is the ChatGPT API cheaper than a ChatGPT Plus plan?
The two bill differently. A plan is a flat monthly fee for a person who chats, and the API charges per token for software that calls a model. Estimate your tokens first, then compare the monthly total with the plan price.
In short: which AI API costs the least?
- Lowest price per token: GPT-6 Luna and Claude Haiku 5.5 at $0.10 input and $0.50 output.
- Lowest cached input: DeepSeek V4.1 Flash at $0.003 per million tokens off-peak.
- Everyday quality: GPT-6.1 Sol and Claude Sonnet 5.5 at $2 input and $10 output.
- Hardest work: GPT-6 Astra and Claude Fable 5.1 at $10 input and $50 output.
Prices are as of 23 October 2026 and change often, so confirm the current rate with the provider before you commit. Gemini 4 Argon doubles on 1 December 2026, and Gemini 3.8 Flash doubles on 1 January 2027.
Add us as a preferred source on Google, and our reviews and picks show up in your Top Stories and AI Overviews.
