Where do the prices come from, and how accurate are they?

All prices are vendor list prices (USD per 1M tokens) from official pricing pages, cross-checked across six third-party sources; the data date is shown in the page footer. We track price changes, but always confirm against the vendor's page before committing.

Why does my real bill differ from the estimate?

The calculator uses the standard non-cached tier. Real bills are also affected by cached-input discounts (as low as 1/10 on some vendors), batch discounts, tiered pricing, FX and taxes — so treat the numbers for model comparison, not as an exact bill forecast.

How is the monthly cost computed?

Monthly cost = monthly input tokens ÷ 1M × input price + monthly output tokens ÷ 1M × output price. For example, 30M input + 10M output tokens on a model priced $1.4 in / $4.4 out per 1M tokens costs 30×1.4 + 10×4.4 = $86 per month.

What does the "runs locally" badge mean?

Badged models ship open weights (e.g. Llama 4 Scout, Qwen3-32B, Gemma 3); the table shows their hosted-endpoint prices. If you have your own GPU, the companion local-hardware matcher can size a self-hosted setup — often cheaper than API calls.

💰

LLM API Cost Calculator

Enter your usage, compare monthly API costs across models.

Free · no sign-up100% in-browser computationText never leaves your device

Your usage

Monthly input tokens30 million

Monthly output tokens10 million

Or pick a typical usage tier:

Monthly cost ranking, all models

Lowest monthly

$6.00

Llama 4 Scout · 30M→10M

Highest monthly

$450.0

GPT-5.6 Sol

Per day (cheapest tier)

$0.200

#	Model	Monthly (USD)	Input / output $/M	Context
1	MetaLlama 4 Scoutcheapestruns locally	$6.00	$0.100 / $0.300	10M
2	GoogleGemma 3 27Bruns locally	$6.00	$0.100 / $0.300	128K
3	DeepSeekDeepSeek V4 Flash缓存命中输入价更低（官方约 $0.03 级）	$7.00	$0.140 / $0.280	1M
4	AlibabaQwen3 32Bruns locally	$7.00	$0.100 / $0.400	32.8K
5	Z.aiGLM-5.3-Flash	$7.60	$0.190 / $0.190	1M
6	xAIGrok 4.1 Fast	$11.00	$0.200 / $0.500	2M
7	GoogleGemini 2.5 Flash	$12.75	$0.175 / $0.750	1M
8	MiniMaxMiniMax M3	$21.00	$0.300 / $1.20	1M
9	GoogleGemini 3 Flash	$45.00	$0.500 / $3.00	1M
10	AnthropicClaude Haiku 4.5	$80.00	$1.00 / $5.00	200K
11	Z.aiGLM-5.2	$86.00	$1.40 / $4.40	1M
12	DeepSeekDeepSeek V4 Pro	$87.00	$1.74 / $3.48	1M
13	xAIGrok 4.6	$120.0	$2.00 / $6.00	500K
14	AlibabaQwen 3.7 Max	$150.0	$2.50 / $7.50	1M
15	GoogleGemini 3 Pro输入超 200K 后约 2×附加	$180.0	$2.00 / $12.00	1M
16	OpenAIGPT-5.6 Terra	$225.0	$2.50 / $15.00	1.1M
17	AnthropicClaude Sonnet 4.6	$240.0	$3.00 / $15.00	1M
18	Moonshot AIKimi K3	$240.0	$3.00 / $15.00	1M
19	AnthropicClaude Opus 4.6全 1M 窗口一口价，无超长上下文附加费	$400.0	$5.00 / $25.00	1M
20	OpenAIGPT-5.6 Sol	$450.0	$5.00 / $30.00	1.1M

List prices from each vendor (USD per 1M tokens, non-cached tier), cross-checked across 6 sources, data as of 2026-09-15. NOT included: cached-input discounts, batch discounts, FX and taxes. All math runs locally in your browser.

工具加载中…

Done calculating? Check these tools

Groq

Groq provides the fastest AI inference in the world with their LPU chips.…

Read review →Website ↗

Portkey

Portkey is an AI development tool designed to streamline workflows and boost…

Read review →Website ↗

Replicate

Replicate makes it easy to run machine learning models in the cloud. Run…

Read review →Website ↗

Related categories

🧩 AI Platform 💻 AI Coding

About LLM API Cost Calculator

Enter your monthly input/output token usage and instantly rank the API cost of 20 major LLMs — from GPT, Claude and Gemini to DeepSeek, GLM and Qwen — at each vendor's list price. Typical usage presets cover side projects to heavy agent workloads, and the result copies as a Markdown table for plan reviews. Pricing is cross-checked across six sources; every calculation runs locally in your browser.

FAQ

Where do the prices come from, and how accurate are they?: All prices are vendor list prices (USD per 1M tokens) from official pricing pages, cross-checked across six third-party sources; the data date is shown in the page footer. We track price changes, but always confirm against the vendor's page before committing.
Why does my real bill differ from the estimate?: The calculator uses the standard non-cached tier. Real bills are also affected by cached-input discounts (as low as 1/10 on some vendors), batch discounts, tiered pricing, FX and taxes — so treat the numbers for model comparison, not as an exact bill forecast.
How is the monthly cost computed?: Monthly cost = monthly input tokens ÷ 1M × input price + monthly output tokens ÷ 1M × output price. For example, 30M input + 10M output tokens on a model priced $1.4 in / $4.4 out per 1M tokens costs 30×1.4 + 10×4.4 = $86 per month.
What does the "runs locally" badge mean?: Badged models ship open weights (e.g. Llama 4 Scout, Qwen3-32B, Gemma 3); the table shows their hosted-endpoint prices. If you have your own GPU, the companion local-hardware matcher can size a self-hosted setup — often cheaper than API calls.

← Browse more free tools