At 2,000 input and 500 output tokens a request, 100,000 requests a month on GPT-6.1 Sol cost $900.00. This free LLM cost calculator prices any workload per request, per month and per year on ten OpenAI, Anthropic and Google models. It runs in your browser, and nothing you enter reaches our server.
Cost per month
$900.00
100,000 requests a month on GPT-6.1 Sol cost about $900.00.
- Cost per month
- $900.00
- Cost per request
- $0.009
- Cost per 1,000 requests
- $9.00
- Cost per year
- $10,800.00
- Share of cost from output
- 56%
- Rates applied
- Standard rates
- Model
- GPT-6.1 Sol
Saved scenarios
Saved in this browser only. Each column is recalculated whenever the page loads.
Every model at this workload, cheapest first
| Model | Vendor | Per 1,000 requests | Per month |
|---|
How this works
- How it works
- Prices are each vendor's published pay-as-you-go API rates from OpenAI's API pricing (opens in new window), Anthropic's Claude pricing (opens in new window) and Google's Gemini API pricing (opens in new window), in US dollars per million tokens, checked on the date shown: OpenAI's and Google's short-context rates switch to long-context rates for the whole request above 272K input tokens (OpenAI) or 200k (Gemini 3.1 Pro), and Claude Haiku 5.5's above 100,000 prompt tokens, a count that includes cache reads; the other Claude models bill one rate across the full context window. Google lists Gemini 3.8 Flash at $0.75 input and $3.75 output through December 31, 2026, doubling on January 1, 2027. Gemini 4 Argon is listed at $2 input and $10 output through November 30, 2026, and $4 and $20 from December 1, 2026. For these two models the calculator applies the price in force on the day you run it, by your device's date, and says which under Rates applied. Each vendor counts tokens with its own tokenizer, so the same text is a different number of tokens on each; Anthropic says its newer tokenizer produces about 30% more tokens than its previous one. The token counter on this site counts with OpenAI's encodings. Daily or monthly totals are billed as one month of tokens, a day being one thirtieth of a month; the typical input per request decides the long-context switch and sets the request count behind the per-request figures.
- Not covered
- Leaves out cache writes and storage, tool and search fees, free tiers, regional uplifts, fast modes and negotiated discounts.
- Data checked
- : Every model price checked against the OpenAI, Anthropic and Google API pricing pages. Gemini 4 Argon's price doubles on Dec 1, 2026 and Gemini 3.8 Flash's on Jan 1, 2027, and the calculator switches on those dates.
- Privacy
- Your numbers never reach our server. The math runs on your device, inside this page.
- Corrections
- Report an error in the math to media@sqmagazine.co.uk; every report is reviewed and confirmed errors are corrected. Include the inputs and the answer you expected. With your permission, we thank you by name on this page.
- Updates
- Gemini 3.5 Flash-Lite's batch cached-input price corrected to $0.015 per million tokens, and Gemini 4 Argon and Gemini 3.8 Flash switch to their announced prices on their effective dates.
- Gemini 4 Argon added; every other model price re-checked against the three vendor pages, unchanged.
- Claude Sonnet 5.5 cache hits now cost $0.10 per million tokens ($0.05 in batch), down from $0.20, and Claude Haiku 5.5 is added with its higher rates for prompts over 100,000 tokens.
- Token counts can be entered as daily or monthly totals, not only per request.
- Feedback
- How useful was this tool?
Table of Contents
How to Use the LLM Cost Calculator
The LLM cost calculator takes one request’s token counts and a monthly volume, then multiplies them by the vendor’s published rates. Six controls set the workload:
- Pick a “Model”. The list holds 10 models: three from OpenAI, four from Anthropic and three from Google.
- Enter “Input tokens per request”: the prompt, system message, chat history and any documents sent with each call.
- Enter “Output tokens per request (including reasoning)”. Anthropic’s thinking guide says the tokens Claude spends reasoning are billed as output tokens, even when the thinking text isn’t returned to you, so count them here.
- Enter “Requests per month”.
- Set “Input served from cache” to the share of input tokens that are cache hits. Leave it at 0 if you don’t use prompt caching.
- Choose “Processing”: “Standard (real time)” or “Batch (results within 24 hours)”.
The panel then shows “Cost per month”, “Cost per request”, “Cost per 1,000 requests”, “Cost per year”, “Share of cost from output” and “Rates applied”. A chart below splits each model from the same vendor into input and output cost, and a table prices all ten models at your workload.
The “Examples” chips load four workloads: a chatbot, retrieval over documents, batch summaries and an agent run. “Copy all details” copies the result panel, and “Save scenario” keeps the inputs in this browser for later. “Share result” copies a link that carries your inputs, and “Download CSV” saves the ten-model table.
Behind every figure in the panel is one short formula.
How LLM API Cost Is Calculated
An LLM API request costs its input tokens times the input rate plus its output tokens times the output rate. Vendors quote each rate per million tokens. OpenAI’s API pricing page lists GPT-6.1 Sol at $2.00 for input and $10.00 for output.
Let I be input tokens, O output tokens, c the cached share and R requests per month. The rates r_in, r_cache and r_out are dollars per million tokens:
- Input cost = I × ((1 – c) × r_in + c × r_cache) ÷ 1,000,000.
- Output cost = O × r_out ÷ 1,000,000.
- Cost per request = input cost + output cost.
- Cost per month = cost per request × R, and cost per year = cost per month × 12.
- Share of cost from output = output cost ÷ cost per request.
Batch mode swaps in the vendor’s batch rates. These are stored in the tool as published, not derived from a discount, because the discount doesn’t apply evenly to every cached-input price.
One rule changes the rates for a whole request. OpenAI lists separate short- and long-context rates for its flagship models.
Its GPT-6.1 Sol page prices prompts with more than 272,000 input tokens at the higher ones. The tool applies that line to all three OpenAI models. Google’s Gemini 3.1 Pro Preview splits its rates at prompts <= 200k tokens. The tool applies both thresholds per request, and "Rates applied" says when a long-context rate is in use.
The pinned workload shows each step of the formula.
LLM Cost Example: 100,000 Requests on GPT-6.1 Sol
The OpenAI API cost calculator example uses the tool’s defaults. That means GPT-6.1 Sol, 2,000 input tokens, 500 output tokens, 100,000 requests a month, no caching and “Standard (real time)”. Six steps rebuild the panel:
- Input: 2,000 tokens at $2.00 per million tokens = $0.004.
- Output: 500 tokens at $10.00 per million tokens = $0.005.
- “Cost per request”: $0.009.
- “Cost per month”: $0.009 × 100,000 requests = $900.00.
- “Cost per year”: $900.00 × 12 = $10,800.00.
- “Share of cost from output”: $0.005 ÷ $0.009 = 56%, and “Rates applied” reads “Standard rates”.
The panel’s “Cost per 1,000 requests” line reads $9.00.
Swap in the other OpenAI models and the same workload runs from $45.00 a month on GPT-6 Luna to $4,500.00 on GPT-6 Astra. Those are list rates, not a measure of which model does the job.
What the Result Means
The figure is an estimate of a pay-as-you-go bill at list rates. Output often carries more of that bill than its token count suggests.
In the example, output is 500 of 2,500 tokens, only 20% of the total. Yet it makes up 56% of the cost, because Sol’s output rate is 5 times its input rate. Tripling the reply to 1,500 tokens lifts the month to $1,900.00.
Reasoning models raise output further. Google’s price list labels the Gemini rate Output price (including thinking tokens), and Anthropic bills Claude’s thinking as output. A reply that shows 500 tokens can bill for far more once the model thinks before it answers.
The tool leaves out cache writes and storage, tool and search fees, free tiers, regional uplifts, fast modes and negotiated discounts. Real traffic also varies from request to request, while the tool assumes every request is the same size.
Treat the monthly figure as a planning number. Run the token counts from a sample of real requests through it before you budget.
API Rates Behind the Calculator
Every rate in the tool comes from the vendor’s own pay-as-you-go price list: OpenAI’s API pricing and Anthropic’s Claude pricing for those two vendors’ models.
Google’s rates come from its Gemini API pricing page. The table lists the Standard rates per million tokens, and every figure matched the vendor pages we captured on October 3.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Astra | $10.00 | $1.00 | $50.00 |
| GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 |
| Claude Fable 5.1 | $10 | $0.25 | $50 |
| Claude Opus 5.5 | $4 | $0.20 | $20 |
| Claude Sonnet 5.5 | $2 | $0.20 | $10 |
| Claude Haiku 4.5 | $1 | $0.10 | $5 |
| Gemini 3.1 Pro Preview | $2.00 | $0.20 | $12.00 |
| Gemini 3.8 Flash | $0.75 | $0.075 | $3.75 |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 |
Source: OpenAI API pricing, Anthropic Claude pricing and Google Gemini API pricing, US dollars per million tokens, Standard tier, short-context rates
Cached input means a cache hit, and Gemini 3.1 Pro Preview’s row covers prompts up to 200,000 tokens. The tool’s registry was checked against the same pages a day earlier, and no rate had moved.
Rates also move on a schedule. Google lists Gemini 3.8 Flash input at $0.75 through December 31, 2026. $1.50 starting January 1, 2027. Output doubles the same day, so the example workload on that model goes from $337.50 to $675.00 a month. Usage and revenue across the three vendors are tracked in our ChatGPT, Claude and Gemini statistics.
How Caching and Batch Processing Change the Bill
Prompt caching lowers the price of repeated input. Batch processing halves the rates in exchange for waiting. Caching works differently by vendor:
- OpenAI: on GPT-6.1 Sol, Cached input tokens are priced at 5% of the uncached input token rate. With 50% of the example’s input cached, the month falls from $900.00 to $710.00.
- Anthropic: A cache hit costs 10% of the standard input price, with lower multipliers on three models, two of them in the tool. On Claude Opus 5.5, a cache hit costs 5% of the standard input price, or $0.20 per million tokens.
- Google: Gemini 3.8 Flash caches at $0.075 against its $0.75 input rate.
Batch trades speed for price. OpenAI’s Batch API guide offers a 50% cost discount compared to synchronous APIs, and Each batch completes within 24 hours. Anthropic’s Batch API carries a 50% discount on both input and output tokens. Google’s Gemini batch documentation prices batch at 50% of the standard cost with a target turnaround time is 24 hours.
The example on Batch costs $450.00 a month. Batch cache prices do not always halve, and the tool stores each one as listed. Gemini 3.1 Pro Preview’s batch cache rate is listed as “Same as Standard”, so its cache hits stay at $0.20 in Batch mode.
Agents resend long, repeated context on every step, which is where caching pays. Their adoption is tracked in our AI coding statistics.
The Long-Context Price Switch
A long prompt can change the price of every token in a request, not only the tokens past the limit. OpenAI’s GPT-6.1 Sol model page says prompts over the threshold are priced at 2x input and cache rates and 1.5x output for the full request. The tool follows that rule exactly.
| Request on GPT-6.1 Sol | Input tokens | Output tokens | Cost per request |
|---|---|---|---|
| At the threshold | 272,000 | 500 | $0.549 |
| One token over | 272,001 | 500 | $1.0955 |
Source: rates from OpenAI’s GPT-6.1 Sol model page and API pricing; costs computed by the tool’s formula
One extra token raises the request’s cost by 99.5%. At 1,000 such requests a month, that is $549.00 against $1,095.50.
Claude 4.6 and later models have no such switch. Anthropic says A 900k-token request is billed at the same per-token rate as a 9k-token request.
Claude Sonnet 5.5 lists the same $2 input and $10 output rates as Sol, so the 272,001-token request still costs $0.549 there. Gemini 3.1 Pro Preview has its own switch: $4.00, prompts > 200k tokens for input. Model choice for coding involves more than price, as our Claude vs ChatGPT for coding comparison shows.
Why Token Counts Differ Between Vendors
The same text is a different number of tokens on each vendor, so a per-token rate is not a per-word price. Each model family splits text with its own tokenizer. Google’s token guide says that For Gemini models, a token is equivalent to about 4 characters. 100 tokens is equal to about 60-80 English words.
Anthropic’s own models differ from each other too. Its pricing page says Claude 4.7 and later models use a newer tokenizer that produces approximately 30% more tokens for the same text. It adds that The exact increase depends on the content and workload shape. Its token-counting guide warns against reusing counts from older models to estimate costs or context window fit.
Two models with equal rates can therefore bill different amounts for the same prompt. Count a sample prompt with each vendor’s own counter, then enter that vendor’s count. The token counter linked under the tool counts with OpenAI’s encodings only.
Do reasoning tokens cost extra?
On Claude and Gemini, reasoning tokens cost the output rate, not a separate fee. Anthropic says thinking tokens count toward `max_tokens` alongside the response text and bill as output, and Google’s output price includes thinking tokens. The extra cost comes from volume. A model that reasons at length produces more output tokens per request.
That’s why the field reads “Output tokens per request (including reasoning)”. On Sol, every 1,000 extra output tokens a request adds $1,000.00 a month at 100,000 requests.
Good to Know Before You Use It
The LLM cost calculator runs in your browser, and nothing you enter reaches our server. It assumes every request carries the same input and output token counts. It prices them at each vendor’s published pay-as-you-go rates in US dollars, checked on the date the tool shows.
The method is the vendor’s per-token arithmetic: input, cached input and output tokens times their rates. OpenAI’s flagship models switch to long-context rates above 272,000 input tokens, the line OpenAI’s GPT-6.1 Sol page states, and Gemini 3.1 Pro Preview above 200,000. Cache writes and storage, tool and search fees, free tiers, regional uplifts, fast modes and negotiated discounts sit outside it. Token counts differ by vendor, so the same text can cost a different amount on each.
SQ Magazine publishes this page for information. The figures are estimates from published list rates, which change. The page promises no security outcome and no final bill; the vendor’s invoice governs what you pay.
Conclusion
At the defaults, 100,000 requests a month on GPT-6.1 Sol cost $900.00, with 56% of it from output. Enter your own token counts, try the same workload on Batch and with caching, and keep inputs under each long-context threshold where you can. Spending across the wider AI market is tracked in our AI market statistics.
