Salary / Tokens.

Your salary, in tokens.

Drag providers and turn your employment cost into tokens.

$3,333.33 / month

+ %
$3,333 / month
$40,000 / year

Gross salary only. Enable to add your employer costs.

AI providers

Drag to add. Click to add or remove.

0 of 4

Budget allocation

Up to four providers. Always adds up to 100%.

Drag a provider onto the bar.

You can also click its icon to select it.

Token capacity · TCE

Pricing checked ·

0 M

tokens / month

0 M

tokens / year

Tokens by provider
Provider Tokens / month Tokens / year

1 M = one million tokens. Month = year ÷ 12.

Settings Input / output 80 / 20 · Cache 0%

The rest are output tokens, including billed reasoning.

ECB · Oct 9, 2026 · editable

Share of input already in the cache.

Reads + writes cannot exceed 100%.

For models that publish both write prices.

Per new token stored, where an hourly rate applies.

Uses the context tiers published in the catalog.

One trillion equals one million million tokens.

Standard text API, excluding tax, subscriptions and extra tools. Unverified cache pricing: standard input rate.

How it works

We start with your gross salary. Enable employer costs to add your chosen percentage; otherwise, we use 0%.

Annual cost = annual gross salary × (1 + employer costs)

Monthly cost is always annual cost divided by 12. If you enter a monthly salary, choose 12 or 14 payments:

We convert EUR budgets to USD. USD budgets are used directly. We allocate the budget to providers, then to their models.

Tokens = model budget ÷ weighted price × 1,000,000

The weighted price accounts for input, output, cache reads and writes. Where storage is billed, we add new stored tokens × hours × hourly rate. Each model is calculated separately, then their tokens are added.

Employment cost reference

Spain’s INE reports monthly employment costs of 3.388,15 € and salary costs of 2.512,68 € for Q2 2026. The difference is approximately 34.84% of salary costs. This is an optional Spanish reference, not a global estimate. Employer costs are off by default in English. Enable them and enter the percentage for your country or company.

View the INE survey

Exchange rate and scope

We fetch the latest published ECB EUR/USD rate through Frankfurter, with a daily cache. Weekends and holidays may use the last business day. You can override the rate in Settings. USD budgets need no conversion. Annual estimates use the selected pricing for 12 months.

View the ECB exchange rate

Calculation happens in your browser. Your salary is never sent to the exchange-rate service or AI APIs.

Why this exists

A KPI proposed by Salary / Tokens

Token Capacity per Employee TCE

Companies measure revenue per employee, workforce costs and team utilization. These figures describe different aspects of an investment: what it costs and what capacity it funds.

Revenue per employee
Revenue ÷ average full-time-equivalent workforce.
Average employment cost
Personnel costs ÷ average workforce, using the same period and scope.
Time utilization
Billable or productive hours ÷ available hours, following each company’s definition.

Bechtle reports revenue per employee and Accenture reports utilization in their 2025 annual reports. Spain’s INE publishes labor costs and effective hours. Companies do not all publish the three indicators or use identical definitions.

Bechtle 2025 ↗ Accenture 2025 ↗ INE 2026 ↗

The question we add

How many AI tokens could one employee’s annual employment cost fund, given a specific mix of providers, models and usage? TCE expresses that purchasing capacity in tokens per employee and period.

TCE = Σ ( each model’s budget ÷ cost per token)

The division is simple. Each token’s price depends on input and output, cache, storage, context, currency and current pricing. We allocate the money first, then calculate each model’s tokens.

See the full formula
TCE year = 10⁶ · S · (1 + e) · x · Σp ωp · Σm (νpm / cpm)
cpm = α [(1 − h − w) Ipm + h Rpm + w Wpm + w τ Apm] + (1 − α) Opm
S, e, x
Annual gross salary, overhead as a fraction (0 if employer costs are disabled), and conversion from the selected currency to USD. In USD, x = 1.
ω, ν
Budget fractions for each provider and its models. Each allocation sums to 1.
α, h, w
Input fraction; fractions of that input read from and written to cache. h + w ≤ 1.
I, R, W, O
USD per million standard input, cache-read, cache-write and output tokens. W uses the selected write duration.
τ, A, c
Storage hours, rate per million stored tokens per hour, and weighted cost per million. Rates account for context, date and peak/off-peak periods where applicable.

Monthly TCE = annual TCE ÷ 12. Without verified cache pricing, standard input is used: h = w = 0. Without a published storage charge, A = 0 in this estimate. If c = 0, the budget does not set a token limit.

How to read the KPI

Use it to compare AI budgets and model mixes. Assessing productivity requires measuring completed tasks, quality, time saved and supervision. Tokens from different models do not have equivalent quality.

Compare using the same cost scope, mix and assumptions. The INE percentage is an editable Spanish reference, including when entering dollars.

Pricing and sources

USD per million tokens · checked Oct 11, 2026

Standard API pricing from the original provider. Expand a provider for models and terms. “Unverified” never means free.

OpenAI 3 models

Standard API, short context. Long context has different pricing.

Pricing for OpenAI, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
GPT-6.1 Sol 2 0.1 2.5 Same as 5 min No published charge 10
GPT-6 Astra 10 1 12.5 Same as 5 min No published charge 50
GPT-6 Luna 0.1 0.01 0.125 Same as 5 min No published charge 0.5
Official source
Anthropic 6 models

Standard API. Cache writes for 5 minutes or 1 hour; reads billed separately.

Pricing for Anthropic, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Claude Sonnet 5.5 2 0.1 2.5 4 No published charge 10
Claude Opus 5.5 4 0.2 5 8 No published charge 20
Claude Haiku 5.5 0.1 0.01 0.125 0.2 No published charge 0.5
Claude Fable 5.1 10 0.25 12.5 20 No published charge 50
Claude Fable 5 10 1 12.5 20 No published charge 50
Claude Opus 5 5 0.5 6.25 10 No published charge 25

Claude Haiku 5.5: Up to 100,000 input tokens per request. Higher pricing applies above this threshold.

Claude Fable 5: Previous generation, still listed in the official pricing table.

Claude Opus 5: Previous generation, still listed in the official pricing table.

Official source
Google Gemini 3 models

Paid tier. Cache storage is billed per million tokens per hour.

Pricing for Google Gemini, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Gemini 3.8 Flash 0.75 0.075 Input price Same as 5 min 0.5 3.75
Gemini 3.5 Flash Lite 0.3 0.03 Input price Same as 5 min 1 2.5
Gemini 3.1 Pro Preview 2 0.2 Input price Same as 5 min 4.5 12

Gemini 3.8 Flash: Current rate through Dec 31, 2026; the announced rate applies afterwards.

Gemini 3.1 Pro Preview: Preview. Different pricing above 200,000 input tokens.

Official source
xAI 1 models

Grok 4.7. Higher pricing for 200,000 input tokens or more.

Pricing for xAI, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Grok 4.7 2 0.5 Input price Same as 5 min No published charge 6
Official source
DeepSeek 2 models

Peak pricing by default. Off-peak: 50% less. Peak: Mon–Fri, 01–04 and 06–10 UTC, excluding Chinese holidays.

Pricing for DeepSeek, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
DeepSeek V4.1 Flash 0.3 0.006 Input price Same as 5 min No published charge 1.2
DeepSeek V4 Pro 0813 1.32 0.044 Input price Same as 5 min No published charge 3.96
Official source
Mistral 4 models

Standard inference pricing. Large 4 is in preview with promotional pricing.

Pricing for Mistral, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Mistral Medium 3.5 1.5 0.15 Input price Same as 5 min No published charge 7.5
Mistral Small 4 0.15 0.015 Input price Same as 5 min No published charge 0.6
Mistral Large 4 0.68 0.07 Input price Same as 5 min No published charge 2.09
Codestral 25.08 0.3 0.03 Input price Same as 5 min No published charge 0.9

Mistral Large 4: Preview promotion without a published end date. Base pricing: 1.36 / 0.14 / 4.18 USD.

Official source
Moonshot Kimi 4 models

International Kimi API. K3 has separate 5-minute and 1-hour write prices.

Pricing for Moonshot Kimi, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Kimi K3 3 0.3 3 6 No published charge 15
Kimi K2.7 Code 0.95 0.19 Input price Same as 5 min No published charge 4
Kimi K2.7 Code Highspeed 1.9 0.38 Input price Same as 5 min No published charge 8
Kimi K2.6 0.95 0.16 Input price Same as 5 min No published charge 4
Official source
Z.ai 3 models

Temporarily free cache storage, according to the official table.

Pricing for Z.ai, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
GLM 5.3 1.4 0.26 Input price Same as 5 min No published charge 4.4
GLM 5.3 Flash 0.15 0.03 Input price Same as 5 min No published charge 0.5
GLM 5.3 FlashX 0.37 0.075 Input price Same as 5 min No published charge 1.25
Official source
Xiaomi MiMo 3 models

International API in USD. Temporarily free cache writes. Excludes models nearing retirement.

Pricing for Xiaomi MiMo, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
MiMo V2.6 Pro 0.435 0.0036 0 Same as 5 min No published charge 0.87
MiMo V2.6 Flash 0.14 0.0028 0 Same as 5 min No published charge 0.28
MiMo V2.6 Pro Ultraspeed 4.35 0.036 0 Same as 5 min No published charge 8.7
Official source
Alibaba Qwen 4 models

International deployment in Singapore, short-context tiers. Explicit cache where public pricing is available. No alias discounts.

Pricing for Alibaba Qwen, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
Qwen 3.7 Plus 0.4 0.04 0.5 Same as 5 min No published charge 1.6
Qwen 3.8 Max 2 Unverified Input price Same as 5 min No published charge 6
Qwen 3.8 Flash 0.15 Unverified Input price Same as 5 min No published charge 0.47
Qwen 3 Coder Flash 0.3 0.03 0.375 Same as 5 min No published charge 1.5

Qwen 3.7 Plus: Snapshot without alias promotions. Explicit cache: reads 10%, writes 125%. Higher tier above 256,000 tokens.

Qwen 3.8 Max: Cache pricing only visible in the console. Standard input pricing applies until verified.

Qwen 3.8 Flash: Cache pricing only visible in the console. Standard input pricing applies until verified.

Qwen 3 Coder Flash: Explicit cache. Input tiers at 32,000, 128,000 and 256,000 tokens.

Official source
MiniMax 3 models

Standard international API. M3 advertises a permanent 50% reduction; priority costs more.

Pricing for MiniMax, USD per million tokens
Model Input Cache read Write 5 min Write 1 h Storage / h Output
MiniMax M3 0.3 0.06 Input price Same as 5 min No published charge 1.2
MiniMax M2.7 0.3 0.06 0.375 Same as 5 min No published charge 1.2
MiniMax M2.7 Highspeed 0.6 0.06 0.375 Same as 5 min No published charge 2.4
Official source

Table values use the short-context tier and the review date. Settings apply published context tiers, DeepSeek off-peak pricing and Gemini’s announced price change. Pricing needs regular review.

Qwen cache terms