Lowest cost for you
Anthropic: Haiku 5.5
≤100K prompt tokens; >100K is $0.50/$2.50
$2/mo
$24 a year · $0.10/MTok in · $0.50/MTok out
Checked 2026-11-05 · Official pricing for Anthropic (opens in a new tab)
Save money › AI API cost
AI API cost calculator
What will the OpenAI, Claude (Anthropic) and Google Gemini API cost for my usage? Enter monthly input and output tokens (in millions). Every $/MTok comes from the official API pricing page, checked 2026-11-05.
1 MTok = 1 million tokens
Presets
Light chatbot: ~2M input + 0.5M output tokens a month · RAG app: ~10M input + 2M output tokens a month · Heavy agent: ~50M input + 15M output tokens a month
Lowest cost for 10M input + 2M output tokens a month
Anthropic Haiku 5.5: $2/mo
$24 a year, the same as OpenAI GPT-6 Luna. Switch models on a card by changing your token inputs above.
Prices for each model
OpenAI: GPT-6 Luna
Cheapest GPT-6 listed
$2/mo
$24 a year · $0.10/MTok in · $0.50/MTok out
Checked 2026-11-05 · Official pricing for OpenAI (opens in a new tab)
Google: Gemini 3.1 Flash-Lite
Text/image/video input rate
$5.50/mo
$66 a year · $0.25/MTok in · $1.50/MTok out
Checked 2026-11-05 · Official pricing for Google (opens in a new tab)
Google: Gemini 3.5 Flash-Lite
High-volume agentic tasks
$8/mo
$96 a year · $0.30/MTok in · $2.50/MTok out
Checked 2026-11-05 · Official pricing for Google (opens in a new tab)
Google: Gemini 3.8 Flash
Through Dec 31, 2026; then $1.50/$7.50
$15/mo
$180 a year · $0.75/MTok in · $3.75/MTok out
Checked 2026-11-05 · Official pricing for Google (opens in a new tab)
OpenAI: GPT-6.1 Terra
Lower-cost GPT-6.1
$20/mo
$240 a year · $1/MTok in · $5/MTok out
Checked 2026-11-05 · Official pricing for OpenAI (opens in a new tab)
Anthropic: Sonnet 5.5
High-performance coding and agents
$40/mo
$480 a year · $2/MTok in · $10/MTok out
Checked 2026-11-05 · Official pricing for Anthropic (opens in a new tab)
Google: Gemini 4 Argon
Flagship; $2/$10 through Nov 30, 2026, then $4/$20
$40/mo
$480 a year · $2/MTok in · $10/MTok out
Checked 2026-11-05 · Official pricing for Google (opens in a new tab)
OpenAI: GPT-6.1 Sol
Balanced GPT-6.1
$40/mo
$480 a year · $2/MTok in · $10/MTok out
Checked 2026-11-05 · Official pricing for OpenAI (opens in a new tab)
Google: Gemini 3.1 Pro Preview
Prompts ≤200K tokens; >200K is $4/$18
$44/mo
$528 a year · $2/MTok in · $12/MTok out
Checked 2026-11-05 · Official pricing for Google (opens in a new tab)
Anthropic: Opus 5.5
Daily driver for agentic coding
$80/mo
$960 a year · $4/MTok in · $20/MTok out
Checked 2026-11-05 · Official pricing for Anthropic (opens in a new tab)
Anthropic: Fable 5.1
Next-gen long-running agents
$200/mo
$2,400 a year · $10/MTok in · $50/MTok out
Checked 2026-11-05 · Official pricing for Anthropic (opens in a new tab)
OpenAI: GPT-6 Astra
Flagship GPT-6 model
$200/mo
$2,400 a year · $10/MTok in · $50/MTok out
Checked 2026-11-05 · Official pricing for OpenAI (opens in a new tab)
How we work it out
- Prices: copied from each official API pricing page on 2026-11-05, before tax. Nothing is estimated.
- Math: monthly cost = (input MTok × $/MTok in) + (output MTok × $/MTok out).
- Tier: Standard / paid-tier list rates only. Batch, priority, flex, and prompt-cache writes are listed under Left out.
- Promos: where Google shows a promo end date, we use the current promo rate and note the later price on the card.
- Not included: tools (web search, grounding), image/audio specialty rates, committed-use discounts, and tax.
Every model we priced
Open a provider to see each verified input and output rate.
OpenAI: every model we priced
| Model | Input $/MTok | Output $/MTok | Notes |
|---|---|---|---|
| GPT-6 Astra | $10 | $50 | Flagship GPT-6 model |
| GPT-6.1 Sol | $2 | $10 | Balanced GPT-6.1 |
| GPT-6.1 Terra | $1 | $5 | Lower-cost GPT-6.1 |
| GPT-6 Luna | $0.10 | $0.50 | Cheapest GPT-6 listed |
Standard tier from OpenAI's API pricing docs. Cache writes and Fast/Ultrafast tiers are not used in the monthly estimate. Checked 2026-11-05 on platform.openai.com (opens in a new tab).
Anthropic: every model we priced
| Model | Input $/MTok | Output $/MTok | Notes |
|---|---|---|---|
| Fable 5.1 | $10 | $50 | Next-gen long-running agents |
| Opus 5.5 | $4 | $20 | Daily driver for agentic coding |
| Sonnet 5.5 | $2 | $10 | High-performance coding and agents |
| Haiku 5.5 | $0.10 | $0.50 | ≤100K prompt tokens; >100K is $0.50/$2.50 |
Claude Platform API standard rates. Prompt caching and batch (50% off) are not applied in the estimate. Checked 2026-11-05 on anthropic.com (opens in a new tab).
Google: every model we priced
| Model | Input $/MTok | Output $/MTok | Notes |
|---|---|---|---|
| Gemini 4 Argon | $2 | $10 | Flagship; $2/$10 through Nov 30, 2026, then $4/$20 |
| Gemini 3.1 Pro Preview | $2 | $12 | Prompts ≤200K tokens; >200K is $4/$18 |
| Gemini 3.8 Flash | $0.75 | $3.75 | Through Dec 31, 2026; then $1.50/$7.50 |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | High-volume agentic tasks |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | Text/image/video input rate |
Gemini Developer API Paid Tier, Standard. Where the page shows a promo end date, that current promo rate is used. Checked 2026-11-05 on ai.google.dev (opens in a new tab).
Left out
- OpenAI Batch / Fast / Ultrafast tiers: the calculator uses Standard rates only; Batch is half and Fast/Ultrafast are higher on the same docs page.
- OpenAI prompt-cache write fees (30m / 12h): shown on the docs table but not part of the simple input+output monthly estimate.
- Anthropic prompt caching and batch: cache read/write and 50% batch discounts are on the pricing page; we price raw input+output.
- Anthropic Haiku 5.5 >100K prompts: priced separately at $0.50/$2.50; the card uses the ≤100K rate and notes the higher tier.
- Gemini Batch / Flex / Priority / Contributor: Standard paid tier only; other service tiers are on the same pricing page.
- Gemini Live / image / TTS / embedding models: billed differently (per-minute audio, per-image, or embedding rates), not plain chat input+output.
- Azure OpenAI, Amazon Bedrock, Vertex AI list prices: marketplace markups vary by region and commitment; use each cloud's own calculator.
FAQ
What will the OpenAI / Claude / Gemini API cost for my usage?
Multiply your monthly input MTok by each model's input $/MTok, add monthly output MTok × output $/MTok. Example: 10M in + 2M out on Claude Sonnet 5.5 is 10×$2 + 2×$10 = $40/mo. Checked 2026-11-05.
Which API is cheapest right now?
At equal token volume, the lowest Standard rates we verified are OpenAI GPT-6 Luna and Anthropic Haiku 5.5 (≤100K prompts), both $0.10/MTok in and $0.50/MTok out, then Gemini 3.1 Flash-Lite at $0.25/$1.50. Flagship models (GPT-6 Astra, Fable 5.1) are $10/$50. Checked 2026-11-05.
Do these prices include prompt caching or batch discounts?
No. The calculator uses each provider's standard paid input and output token rates only. Anthropic lists prompt-cache read/write and 50% batch; OpenAI lists cache writes and Batch/Fast/Ultrafast; Gemini lists Batch/Flex/Priority. Those can change the bill.
What is an MTok?
One MTok is 1 million tokens. Providers price API use per million input tokens and per million output tokens. A short chat turn is often a few thousand tokens; a RAG pipeline that stuffs big contexts can burn millions of input tokens a month.
Why is openai.com/api/pricing not your OpenAI source?
That URL currently shows ChatGPT Business and Enterprise seat plans, not API token rates. Token rates are on platform.openai.com/docs/pricing, which is what we checked.
Are free tiers included?
Gemini's Free Tier shows free input/output for some Flash models with limits; OpenAI and Anthropic API usage is pay-as-you-go (Claude chat Free is a product plan, not these API rates). The cards show paid Standard rates so you can budget production traffic.
Related: AI coding tool cost calculator · AI usage limits · ChatGPT vs Claude vs Gemini · ChatGPT · Claude · Google Gemini · Save money · How we pick