LLM token cost calculator
What will your AI workload cost on every model?
Describe your traffic once and compare the monthly bill across 330 AI models, cheapest first. Compare cost, published intelligence and value for money for your workload. Narrow the results by the capabilities your app needs.
API calls per month.
Prompt + history + context.
Length of each answer.
150M input + 40M output tokens per month
Must have
Best AI models for your workload
Lowest monthly cost
GLM 5.3 FlashFor 100,000 requests per month with 1,500 input and 400 output tokens each, GLM 5.3 Flash is the cheapest matching model at $8.35/month, with an Intelligence Index of 41.8.
Highest intelligence
Claude Opus 5.5Claude Opus 5.5 leads the matching models on published AA Intelligence at 57.6, with an estimated cost of $1,400/month for the same workload. That is $1,392 more per month than the cheapest option.
Highest intelligence per dollar among models reaching at least 80% of the leading matching score. A benchmark-based value estimate, not a guarantee of quality for every task.
Cheapest models
40$0.0335 in · $0.083 out /1M · 1M context · $0.08/1K req
$0.091 in · $0.182 out /1M · 1M context · $0.21/1K req
$0.051 in · $0.567 out /1M · 1M context · $0.30/1K req
$0.14 in · $0.28 out /1M · 1.1M context · $0.32/1K req
$0.1 in · $0.5 out /1M · 1.1M context · $0.35/1K req
$0.2088 in · $0.4176 out /1M · 1M context · $0.48/1K req
$0.2 in · $1.2 out /1M · 1.1M context · $0.78/1K req
$0.15 in · $1.875 out /1M · 1M context · $0.97/1K req
$0.435 in · $0.87 out /1M · 1.1M context · $1.00/1K req
$0.44 in · $1.32 out /1M · 1M context · $1.19/1K req
Most intelligent models
40$4 in · $20 out /1M · 1M context · $14/1K req
$10 in · $50 out /1M · 1M context · $35/1K req
$10 in · $50 out /1M · 1.1M context · $35/1K req
$5 in · $25 out /1M · 1M context · $18/1K req
$10 in · $50 out /1M · 1M context · $35/1K req
$1.25 in · $4.25 out /1M · 1M context · $3.58/1K req
$2 in · $10 out /1M · 1.1M context · $7.00/1K req
$4 in · $20 out /1M · 1.1M context · $14/1K req
$2 in · $6 out /1M · 500K context · $5.40/1K req
$0.435 in · $0.87 out /1M · 1.1M context · $1.00/1K req
Top 10 cheapest at a glance
Estimated monthly cost in USD · lowest first
Monthly cost ranking data
| Model | USD / month | AA Index |
|---|---|---|
| GLM 5.3 Flash | $8.35 | 41.8 |
| DeepSeek V4 Flash 0423 | $21 | 34.3 |
| DeepSeek V4.1 Flash | $30 | 39.5 |
| MiMo-V2.6-Flash | $32 | 37.9 |
| GPT-6 Luna | $35 | 38.1 |
| DeepSeek V4 Pro 0423 | $48 | 36.0 |
| GPT-5.6 Luna | $78 | 37.3 |
| Qwen3.8 27B | $98 | 33.7 |
| MiMo-V2.6-Pro | $100 | 46.3 |
| DeepSeek V4 Flash Vision Exp | $119 | 34.8 |
Top 10 most intelligent at a glance
Published AA Intelligence Index · highest first
Intelligence ranking data
| Model | USD / month | AA Index |
|---|---|---|
| Claude Opus 5.5 | $1,400 | 57.6 |
| Claude Fable 5.1 | $3,500 | 53.4 |
| GPT-6 Astra | $3,500 | 52.7 |
| Claude Opus 5 | $1,750 | 50.8 |
| Claude Fable 5 | $3,500 | 49.6 |
| Muse Spark 1.3 | $358 | 48.1 |
| GPT-6 Sol | $700 | 47.6 |
| GPT-5.6 Sol | $1,400 | 47.0 |
| Grok 4.7 | $540 | 46.4 |
| MiMo-V2.6-Pro | $100 | 46.3 |
Both charts use the same workload and active filters: 100,000 monthly requests, 1,500 input and 400 output tokens per request. Intelligence rankings include only published AA Intelligence Index scores; missing scores are never estimated.
How LLM token pricing works
AI model APIs bill per token. Everything you send — system prompt, chat history, retrieved documents and the user's message — counts as input. Everything the model writes counts as output, which is usually 3–8× more expensive per token.
monthly cost = requests × (input tokens × input $/1M + output tokens × output $/1M) ÷ 1,000,000
That is why the cheapest model changes with your workload: RAG and agents resend long context and are input-heavy, while chat and writing apps are output-heavy. Pick the closest preset, adjust the numbers, then open any model's own calculator for a detailed breakdown or compare two models side by side.
Prices track each model's listed rates; see the AI Model Price Index for recent changes.
LLM cost calculator FAQ
How do I calculate LLM API cost?+
Monthly cost = requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000. Prices are listed per million tokens, so the calculator multiplies your monthly token totals by each model's input and output rate.
What is the cheapest LLM API?+
It depends on your prompt and answer sizes, because models weigh input and output prices differently. Set your workload above and the list ranks every tracked model from cheapest to most expensive for exactly that mix.
How many tokens is a word?+
In English, one token is about 4 characters or ¾ of a word, so 1,000 tokens is roughly 750 words. Code, non-English text and JSON usually use more tokens per word.
Why filter by features like JSON mode or function calling?+
The cheapest model is only useful if it can do the job. Requiring structured outputs, tool calls, web search, image input or a minimum context window removes models that cannot run your workload.
Do these estimates include caching or batch discounts?+
No. Estimates use standard listed input and output prices. Prompt caching, batch APIs and volume contracts can lower the real bill; taxes and tool fees can raise it.
How is the best-value AI model chosen?+
Among matching models with published Artificial Analysis Intelligence Index scores, we keep those reaching at least 80% of the highest matching score, then select the highest score per dollar of estimated monthly cost. This is a budget-oriented heuristic, not a claim that benchmark points scale linearly with real-world quality. At zero usage no value recommendation is made.
What does the Intelligence Index filter measure?+
The range uses published Artificial Analysis Intelligence Index scores, not estimated scores. It starts at 30–60. Models without a published score only appear when the range is cleared to 0–100 and are never included in the most-intelligent ranking. Selected input and output capabilities must all be listed for a model to match. Costs still use text-token rates; image, audio and video processing may have additional billing.
How often are model prices updated?+
PriceIndex checks model prices several times a day and records every change in the AI Model Price Index.