AI Token Calculator: Count Tokens & Estimate Cost

Count text tokens or estimate total-volume and per-request API costs across 126 eligible models.

1 page ≈ 500 words / 750 input tokens; chat assumes 250 output tokens (3:1). Editable example; counts vary by model and language. K / M / B shortcuts supported. Total volume is not a single prompt.

$0.008 estimated total

anthropic

Standard-rate estimate for total volume. Prompt sizes are unknown; long-context requests may cost more.

Advanced: cache & hidden reasoning

Reasoning tokens are billed as output tokens, are not shown in the response, and vary strongly by reasoning effort and task length. Check your own API usage (reasoning_tokens in the usage object) for a real number. Enter visible output separately above to avoid double-counting; each estimate adds your hidden-token count once at that model’s output rate. See our reasoning-effort experiment.

Scheduled billing terms by offer

DeepSeek V4.1 Flash · deepseek: Token estimates use the off-peak reference, not an all-hours quote. Peak: $0.3 input / $0.006 cached input / $1.2 output per 1M tokens during 01:00-04:00 and 06:00-10:00 UTC, on Monday through Friday. Weekends and Chinese public holidays are off-peak in full. Separate per-request tool fees are excluded.

DeepSeek V4 Pro 0813 · deepseek: Token estimates use the off-peak reference, not an all-hours quote. Peak: $1.32 input / $0.044 cached input / $3.96 output per 1M tokens during 01:00-04:00 and 06:00-10:00 UTC, on Monday through Friday. Weekends and Chinese public holidays are off-peak in full. Separate per-request tool fees are excluded.

Comparing a subscription? Compare plans or find your subscription vs API break-even.

Advanced model filters
ClassAccess
Provider
  • Claude Opus 5.5
    AnthropicInput $0.003Output $0.005
    Total $0.008
  • Gemini 4 Argon
    GoogleInput $0.0015Output $0.0025
    Total $0.004
  • DeepSeek V4.1 Flash
    DeepSeekInput $0.000113Output $0.00015
    Total $0.000263
  • GPT-6 Astra
    OpenAIInput $0.0075Output $0.0125
    Total $0.02
  • Gemini 4 Carbon
    GoogleInput $0.0015Output $0.0025
    Total $0.004
  • GPT-6.1 Sol
    OpenAIInput $0.0015Output $0.0025
    Total $0.004
  • Grok 4.7
    xAIInput $0.0015Output $0.0015
    Total $0.003
  • GLM-5.3
    Z.aiInput $0.00105Output $0.0011
    Total $0.00215
  • Grok 4.6
    xAIInput $0.0015Output $0.0015
    Total $0.003
  • Kimi K3
    MoonshotInput $0.00225Output $0.00375
    Total $0.006
  • GLM-5.2
    Z.aiInput $0.00105Output $0.0011
    Total $0.00215
  • DeepSeek V4 Pro 0813
    DeepSeekInput $0.000495Output $0.000495
    Total $0.00099
  • Qwen3.7-Max
    AlibabaInput $0.001875Output $0.001875
    Total $0.00375
  • Kimi K3.1 Pro
    NovitaInput $0.0015Output $0.0025
    Total $0.004
  • Kimi K3.1 Pro Highspeed
    NovitaInput $0.003Output $0.005
    Total $0.008
  • DeepSeek V4 Pro 0813
    NovitaInput $0.00099Output $0.00099
    Total $0.00198
  • Qwen3.8 Max
    NovitaInput $0.0015Output $0.0015
    Total $0.003
  • Kimi K3
    telnyxInput $0.002025Output $0.003375
    Total $0.0054
  • Qwen3 235B A22B Instruct 2507
    TogetherInput $0.00015Output $0.00015
    Total $0.0003
  • MiniMax M3
    TogetherInput $0.000225Output $0.0003
    Total $0.000525
126 models · 750 in · 250 out

How much do AI tokens cost?

Listed standard rates range from $0.02–$10 per million input tokens and $0.05–$50 per million output tokens. These are separate rates, not a blended price. Unavailable and historical offers are excluded. Source verification varies by row; confirm provider terms.

Input + output volumeStandard-rate cost range
100,000 each$0.007–$6
1,000,000 each$0.07–$60
1,000,000,000 each$70–$60,000

Pricing data: How we verify →

Want input, cached and output rates side by side without a workload? Browse the API price table. Comparing metered APIs with your own hardware or a rented GPU? Run the local AI vs API calculator.

126 models · 15 providers Data updated 2026-11-12 Per-model verification dates

How do I use the AI token cost calculator?

  1. Enter expected input tokens — choose total volume for a period or input length per request. Count tokens uses a GPT tokenizer with a labelled character fallback; counts vary by model and language.
  2. Enter expected output tokens — enter total output for the period or visible output per request; add hidden reasoning separately under Advanced.
  3. Set monthly request volume — in per-request mode only; prompt thresholds use each request, not monthly volume. Total-volume mode uses standard rates and may underestimate long-context requests.
  4. Compare rows — models can be filtered by provider, tier and access route; switch to lowest-cost sorting whenever price is the priority.
  5. Review model details for verification dates and official sources; confirm current provider pricing before purchasing.

Methodology

Rates come from the same maintained dataset as the API pricing table (last update 2026-11-12). Per-model verification dates and official sources remain in the dataset and model details. The provider page is authoritative if pricing changes. Total-volume mode uses standard rates without treating the period total as one prompt. Per-request mode selects the listed model-specific context rates using input length per request, then multiplies the following cost by requests per month; source-review warnings still apply:

cost = (input_tokens / 1,000,000 * input_rate) + (output_tokens / 1,000,000 * output_rate)

When the cached-input slider is above 0%, the input portion splits into cached and non-cached fractions, each multiplied by the respective rate. Models without a published cached rate use the standard input rate for both.

Only priced rows with an official source and a recorded verification date are published; a priced row without provenance fails the build. A recorded date does not guarantee current verification: unresolved or stale source evidence remains explicitly warned.

Estimates do not validate provider context, output, account or rate limits; check the provider’s official documentation before use.

Billing scope: listed synchronous API offers, identified by provider and model ID, in USD per million tokens. Hosted offers use their listed provider rates; these are not a marketplace invoice estimate. Excludes cache writes/storage, batch/fast/priority tiers, regional modifiers, tools, taxes, subscription entitlements, non-token billing and negotiated discounts.

Frequently asked questions about AI token costs

How do I calculate AI API costs?

Multiply input tokens by the input rate and output tokens by the output rate, divide each by 1,000,000, then add the two amounts. Choose total-volume standard estimates or per-request pricing multiplied by requests per month; context thresholds apply only per request. Advanced includes cached input and manually entered hidden reasoning.

Which API models are included?

This calculator retains 272 canonical catalog entries from the same dataset as the API pricing table, including 126 non-historical entries. Estimates cover 126 eligible listed offers from 15 providers. Unavailable prices are excluded from estimates; legacy and deprecated entries are hidden by default and can be shown with the legacy toggle. Recorded verification dates, official sources and source-review states remain in the maintained dataset and model details.

What is the lowest listed output rate shown?

Among the 126 eligible listed models, Llama 3.1 8B Instruct has the lowest listed output rate at $0.05 per million output tokens. Confirm current pricing with the provider before purchasing.

Are input and output tokens priced the same?

Not always. Each listed offer has separate input and output rates, with its recorded source-review state. Cached input has its own rate where listed.

How accurate is this calculator?

The dataset was last updated on 2026-11-12; each row keeps its own verification date from maintained official-source data. Provider pages remain authoritative, and rates may change after that date.