AI Tools: every tool in detail

Every tool in AI Tools (43), with how to use it and its real options. Read the category guide for the big picture and the reference list for a quick scan. Options, defaults and worked examples are generated from the code, so they match the tool exactly.

Tokens & context#

Words to Tokens Calculator#

Open tool →Your browser (nothing is uploaded)

Estimate tokens from a word count.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Words1500number

Worked example (with the default values): Tokens (est.) → 2,000

DetailValue
Characters (est.)8,550
Pages (500 words)3

Rule of thumb for English with GPT- and Claude-style tokenisers: about 4 characters or 0.75 words per token. Code and non-English text use more tokens.

Tokens to Words Calculator#

Open tool →Your browser (nothing is uploaded)

How many words fit in a token budget.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tokens8000number

Worked example (with the default values): Words (est.) → 6,000

DetailValue
Characters (est.)32,000
Pages (500 words)12

Rule of thumb for English with GPT- and Claude-style tokenisers: about 4 characters or 0.75 words per token. Code and non-English text use more tokens.

Characters to Tokens Calculator#

Open tool →Your browser (nothing is uploaded)

Estimate tokens from a character count.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Characters12000number

Worked example (with the default values): Tokens (est.) → 3,000

DetailValue
Words (est.)2,105

Rule of thumb for English with GPT- and Claude-style tokenisers: about 4 characters or 0.75 words per token. Code and non-English text use more tokens.

File Size to Tokens Calculator#

Open tool →Your browser (nothing is uploaded)

Roughly how many tokens a text file holds, from its size.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
File size250KB
Content4English prose, Source code, JSON / logs, CJK text (UTF-8)

Worked example (with the default values): Tokens (est.) → 62,500

DetailValue
Bytes per token assumed4
Fits in a 200k context?Yes
Words (est.)46,875

Applies to plain-text formats. PDFs and Office files contain layout data, so their size overstates the text.

Context Window Calculator#

Open tool →Your browser (nothing is uploaded)

What's left for your content after the system prompt, history and reply.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Context window200000tokens
System prompt1800tokens
Conversation history12000tokens
Reserved for the reply4096tokens

Worked example (with the default values): Free for your content → 182,104 tokens

DetailValue
That's about136,578 words · 273 pages
Used8.9%

Rule of thumb for English with GPT- and Claude-style tokenisers: about 4 characters or 0.75 words per token. Code and non-English text use more tokens.

Context Usage Calculator#

Open tool →Your browser (nothing is uploaded)

How full a context window is, and how many more turns fit.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tokens used46000number
Context window128000number
Average tokens per turn900number

Worked example (with the default values): Used → 35.9%

DetailValue
Remaining82,000 tokens
Turns left91

Max Output Length Calculator#

Open tool →Your browser (nothing is uploaded)

Words and pages a max-token setting allows.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Max output tokens4096number

Worked example (with the default values): Up to → 3,072 words

DetailValue
Pages6.1
Speaking time (150 wpm)20.5 min

Rule of thumb for English with GPT- and Claude-style tokenisers: about 4 characters or 0.75 words per token. Code and non-English text use more tokens.

Token Counter#

Open tool →Your browser (nothing is uploaded)

Estimate the tokens in any text, and how much of a context window it uses.

How to use

  1. Paste or type your text into the input box (or arrive here with text via Send to…).
  2. The result appears instantly on the other side.
  3. Copy it, or download if offered.

Example

Input:

The report was written by the committee last week. Honestly, it is very important that we really think outside the box at the end of the day. However, the results were clearly ignored, and mistakes were made.

Short sentences help. Long sentences that wander through several clauses, pile up qualifi…

Output:

Tokens (estimate):      103
Characters:             412
Words:                  69
Characters per token:   4.00
Share of a 8k window:   1.3%
Share of a 128k window: 0.08%
Share of a 200k window: 0.05%

This is an estimate from word and punctuation patterns, usually within about 10% for English.
Eac…

RAG & embeddings#

Chunk Count Calculator#

Open tool →Your browser (nothing is uploaded)

How many chunks a document splits into, with overlap.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Document tokens120000number
Chunk size800tokens
Overlap100tokens

Worked example (with the default values): Chunks → 172

DetailValue
Tokens embedded (with overlap)137,600
Overhead from overlap14.7%

Chunk Size Calculator#

Open tool →Your browser (nothing is uploaded)

The largest chunk size that lets top-k results fit in the prompt.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Context window16000tokens
Prompt and question1200tokens
Reserved for the answer1500tokens
Chunks retrieved (top-k)6number

Worked example (with the default values): Max chunk size → 2,216 tokens

DetailValue
Budget for retrieved text13,300 tokens
About1,663 words per chunk

Knowledge Base Size Calculator#

Open tool →Your browser (nothing is uploaded)

Tokens, chunks and vector storage for a document collection.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Documents5000number
Average pages each12number
Tokens per page650number
Chunk size500tokens
Embedding dimensions1536number

Worked example (with the default values): Chunks → 78,000

DetailValue
Total tokens39,000,000
Vector storage (float32)479.23 MB
With index overhead (~1.5×)718.85 MB

Embedding Cost Calculator#

Open tool →Your browser (nothing is uploaded)

One-off cost to embed a corpus, and the cost to keep it fresh.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tokens to embed40000000number
Price per million tokens0.02$
Re-embedded each month10%

Worked example (with the default values): Initial cost → $0.80

DetailValue
Monthly refresh$0.08
First year$1.76

Vector Database Size Calculator#

Open tool →Your browser (nothing is uploaded)

Memory needed for a vector index.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Vectors2000000number
Dimensions1024number
Precision4float32, float16, int8 (quantised), binary
Metadata per vector200bytes

Worked example (with the default values): Index size → 12.69 GB

DetailValue
Raw vectors8.19 GB
HNSW overhead (~50%)4.1 GB
Metadata400 MB

RAG ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Value of answering questions from your documents against the running cost.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Questions per month6000number
Minutes saved per question6min
Hourly cost of staff45$
Answers that are usable80%
Running cost per month1800$

Worked example (with the default values): Net value per month → $19,800.00

DetailValue
Hours saved480
ROI1,100%
Cost per usable answer$0.38

Cost & pricing#

LLM API Cost Calculator#

Open tool →Your browser (nothing is uploaded)

Cost per request, per day and per month from your token counts and prices.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Input tokens per request2500number
Output tokens per request600number
Input price per million3$
Output price per million15$
Requests per day2000number

Worked example (with the default values): Per month → $990.00

DetailValue
Per request$0.0165
Per day$33.00
Input share45%
Per 1,000 requests$16.50

Enter the current prices from your provider's pricing page: they change often.

Prompt Caching Savings Calculator#

Open tool →Your browser (nothing is uploaded)

What caching a repeated prompt prefix saves.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Cached prefix8000tokens
Requests per day3000number
Input price per million3$
Cached-read discount90%
Cache hit rate85%

Worked example (with the default values): Saved per month → $1,652.40

DetailValue
Without caching$72.00 / day
With caching$16.92 / day
Reduction76.5%

Ignores the one-off cache-write premium some providers charge.

Batch API Savings Calculator#

Open tool →Your browser (nothing is uploaded)

The saving from sending non-urgent jobs through a batch endpoint.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Current monthly spend4000$
Share that can wait40%
Batch discount50%

Worked example (with the default values): Saved per month → $800.00

DetailValue
New monthly spend$3,200.00
Per year$9,600.00

Model Cost Comparison#

Open tool →Your browser (nothing is uploaded)

The same workload priced on two models.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Input tokens per month500M
Output tokens per month80M
Model A input price3$ / M
Model A output price15$ / M
Model B input price0.8$ / M
Model B output price4$ / M

Worked example (with the default values): Model B is cheaper by → $1,980.00

DetailValue
Model A$2,700.00
Model B$720.00
Ratio3.75×

Fine-Tuning Cost Calculator#

Open tool →Your browser (nothing is uploaded)

Training cost from dataset size and epochs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Training examples5000number
Tokens per example800number
Epochs3number
Training price per million tokens8$

Worked example (with the default values): Training cost → $96.00

DetailValue
Tokens trained12,000,000
Per epoch$32.00

AI Infrastructure Cost Calculator#

Open tool →Your browser (nothing is uploaded)

Monthly cost of self-hosting on GPUs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
GPUs4number
Price per GPU-hour2.2$
Hours per day running24number
Storage and networking600$ / mo
Engineering time4000$ / mo

Worked example (with the default values): Per month → $10,936.00

DetailValue
GPU compute$6,336.00
Per year$131,232.00

Self-Hosting vs API Break-Even#

Open tool →Your browser (nothing is uploaded)

The monthly token volume where your own GPUs beat an API.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
API blended price per million tokens2.5$
Self-hosting fixed cost per month9000$
Your tokens per month2500M

Worked example (with the default values): Break-even volume → 3,600 M tokens / mo

DetailValue
API cost at your volume$6,250.00
Self-hosting cost$9,000.00
Cheaper nowAPI

AI Usage-Based Pricing Calculator#

Open tool →Your browser (nothing is uploaded)

Price per unit that keeps your target margin over model costs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Model cost per unit (request, doc, minute)0.018$
Other cost per unit0.004$
Target gross margin75%

Worked example (with the default values): Charge per unit → $0.088

DetailValue
Cost per unit$0.022
Per 1,000 units$88.00
Markup300%

AI SaaS Pricing Calculator#

Open tool →Your browser (nothing is uploaded)

A per-seat price that covers model usage and keeps your margin.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Requests per user per month400number
Model cost per request0.012$
Other cost per user3$
Target gross margin80%

Worked example (with the default values): Price per seat → $39.00 / mo

DetailValue
Cost per seat$7.80
Model share of cost62%
A heavy user (5× usage) costs$27.00

AI Gross Margin Calculator#

Open tool →Your browser (nothing is uploaded)

Gross margin once inference is counted as cost of revenue.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Revenue120000$ / mo
Inference cost26000$ / mo
Hosting and other COGS9000$ / mo

Worked example (with the default values): Gross margin → 70.8%

DetailValue
Gross profit$85,000.00
Inference as % of revenue21.7%
Margin if inference halves81.7%

AI Product MRR Calculator#

Open tool →Your browser (nothing is uploaded)

MRR from seats plus metered usage.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Paying seats850number
Price per seat30$
Usage overage revenue6400$ / mo

Worked example (with the default values): MRR → $31,900.00

DetailValue
ARR$382,800.00
Usage share20%

AI Product LTV Calculator#

Open tool →Your browser (nothing is uploaded)

Customer lifetime value after inference costs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Revenue per customer60$ / mo
Inference cost per customer11$ / mo
Other cost to serve4$ / mo
Monthly churn3.5%

Worked example (with the default values): LTV → $1,285.71

DetailValue
Margin per month$45.00
Lifetime28.6 months
LTV before inference$1,600.00

AI Product CAC Calculator#

Open tool →Your browser (nothing is uploaded)

Acquisition cost including free-tier inference.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Marketing and sales spend40000$
Free users acquired12000number
Inference cost per free user0.4$
Free to paid3%

Worked example (with the default values): CAC → $124.44

DetailValue
New customers360
Free-tier inference$4,800.00
CAC without free-tier cost$111.11

AI Unit Economics Calculator#

Open tool →Your browser (nothing is uploaded)

Profit per request once every cost is counted.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Revenue per request0.05$
Input tokens3000number
Output tokens700number
Input price per million3$
Output price per million15$
Other cost per request0.004$

Worked example (with the default values): Profit per request → $0.0265

DetailValue
Cost per request$0.0235
Margin53%
Break-even price$0.0235

ROI & savings#

AI ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Return on an AI project from time saved against what it costs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tasks per month4000number
Minutes saved on each10min
Loaded hourly cost of staff40$
Share handled by AI60%
AI cost per month2500$

Worked example (with the default values): Net saving per month → $13,500.00

DetailValue
Hours saved400
Labour value$16,000.00
ROI540%
Payback on a $10,000 setup0.7 months
Per year$162,000.00

AI Automation Savings Calculator#

Open tool →Your browser (nothing is uploaded)

Labour saved by automating a repetitive task.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tasks per month10000number
Minutes saved on each4min
Loaded hourly cost of staff32$
Share handled by AI80%
AI cost per month1500$

Worked example (with the default values): Net saving per month → $15,566.67

DetailValue
Hours saved533.3
Labour value$17,066.67
ROI1,038%
Payback on a $10,000 setup0.6 months
Per year$186,800.00

AI Customer Support Savings Calculator#

Open tool →Your browser (nothing is uploaded)

Savings from AI resolving a share of support tickets.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tickets per month8000number
Minutes saved on each9min
Loaded hourly cost of staff30$
Share handled by AI40%
AI cost per month3000$

Worked example (with the default values): Net saving per month → $11,400.00

DetailValue
Hours saved480
Labour value$14,400.00
ROI380%
Payback on a $10,000 setup0.9 months
Per year$136,800.00

AI Chatbot ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Return on a chatbot that deflects conversations from your team.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Conversations per month15000number
Minutes saved on each6min
Loaded hourly cost of staff28$
Share handled by AI50%
AI cost per month2000$

Worked example (with the default values): Net saving per month → $19,000.00

DetailValue
Hours saved750
Labour value$21,000.00
ROI950%
Payback on a $10,000 setup0.5 months
Per year$228,000.00

AI Agent ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Return on an agent that completes whole tasks end to end.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tasks per month1200number
Minutes saved on each35min
Loaded hourly cost of staff55$
Share handled by AI45%
AI cost per month4000$

Worked example (with the default values): Net saving per month → $13,325.00

DetailValue
Hours saved315
Labour value$17,325.00
ROI333%
Payback on a $10,000 setup0.8 months
Per year$159,900.00

AI Productivity Calculator#

Open tool →Your browser (nothing is uploaded)

Value of time an assistant saves a team.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tasks per month6000number
Minutes saved on each8min
Loaded hourly cost of staff60$
Share handled by AI70%
AI cost per month1800$

Worked example (with the default values): Net saving per month → $31,800.00

DetailValue
Hours saved560
Labour value$33,600.00
ROI1,767%
Payback on a $10,000 setup0.3 months
Per year$381,600.00

AI Cost Reduction Calculator#

Open tool →Your browser (nothing is uploaded)

Operating cost removed by handing work to AI.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Work items per month20000number
Minutes saved on each5min
Loaded hourly cost of staff35$
Share handled by AI55%
AI cost per month5000$

Worked example (with the default values): Net saving per month → $27,083.33

DetailValue
Hours saved916.7
Labour value$32,083.33
ROI542%
Payback on a $10,000 setup0.4 months
Per year$325,000.00

AI Workforce Savings Calculator#

Open tool →Your browser (nothing is uploaded)

Hours and cost freed across a workforce.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Tasks per month30000number
Minutes saved on each7min
Loaded hourly cost of staff42$
Share handled by AI35%
AI cost per month9000$

Worked example (with the default values): Net saving per month → $42,450.00

DetailValue
Hours saved1,225
Labour value$51,450.00
ROI472%
Payback on a $10,000 setup0.2 months
Per year$509,400.00

AI Sales Agent ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Extra revenue from faster lead follow-up against the tool's cost.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Leads per month2000number
Current lead-to-sale2%
Lift from instant follow-up20%
Average deal value900$
Cost per month2500$

Worked example (with the default values): Net gain per month → $4,700.00

DetailValue
Extra sales8
Extra revenue$7,200.00
ROI188%

AI Marketing ROI Calculator#

Open tool →Your browser (nothing is uploaded)

Content output and cost per piece with and without AI.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Pieces produced per month40number
Hours per piece without AI5number
Time reduction with AI55%
Hourly cost50$
AI tools per month400$

Worked example (with the default values): Net saving per month → $5,100.00

DetailValue
Cost per piece before$250.00
Cost per piece after$122.50
Extra pieces in the same hours49

AI Revenue Impact Calculator#

Open tool →Your browser (nothing is uploaded)

Revenue change from a conversion or retention lift.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Monthly revenue300000$
Conversion lift4%
Churn reduction0.5points
Current monthly churn4%

Worked example (with the default values): Added revenue per month → $12,000.00

DetailValue
Per year from conversion$144,000.00
Customer lifetime grows25 → 28.6 months
LTV increase14.3%

Hardware & speed#

LLM VRAM Calculator#

Open tool →Your browser (nothing is uploaded)

GPU memory needed to run a model at a given quantisation.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Parameters13billion
Precision44-bit (Q4), 5-bit (Q5), 8-bit (Q8), 16-bit (FP16), 32-bit (FP32)
Context length8192tokens
Layers40number
Hidden size5120number

Worked example (with the default values): VRAM needed → 13.5 GB

DetailValue
Weights6.1 GB
KV cache (FP16)6.25 GB
Fits on16 GB card

Weights = parameters × bits ÷ 8. KV cache = 2 × layers × context × hidden size × 2 bytes. A 10% allowance covers activations and runtime overhead.

Tokens per Second Calculator#

Open tool →Your browser (nothing is uploaded)

Response time from generation speed.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Output tokens800number
Speed60tokens / s
Time to first token0.6s

Worked example (with the default values): Response time → 13.9 s

DetailValue
Words per minute generated2,700
A person reads about250 words per minute

GPU Rental Cost Calculator#

Open tool →Your browser (nothing is uploaded)

Cost of a training or inference job on rented GPUs.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
GPUs8number
Hours36number
Price per GPU-hour2.5$

Worked example (with the default values): Job cost → $720.00

DetailValue
GPU-hours288
If left running a month$14,400.00

Training Time Estimator#

Open tool →Your browser (nothing is uploaded)

Wall-clock time for a training run from throughput.

How to use

  1. Enter the values you know in the fields below (defaults are filled in so you can see how it works).
  2. The result updates as you type; no button needed.
  3. Copy the result or the working lines shown under it.
FieldDefaultChoices / unit
Examples2000000number
Epochs3number
Throughput450examples / s

Worked example (with the default values): Training time → 3.7 hours

DetailValue
Per epoch1.23 hours
Steps at batch 32187,500