Skip to main content
LLM Math Pro

Methodology · Overview

Token Pricing Methodology

Reviewed by · Last reviewed .

Primary sources

Token pricing data is sourced directly from each provider's official pricing documentation: Anthropic (anthropic.com/pricing), OpenAI (openai.com/pricing), Google (ai.google.dev/pricing for Gemini API), and Meta (ai.meta.com for Llama hosted inference). We do not use third-party aggregators — only provider-canonical pricing pages.

Prices are expressed as cost per 1 million tokens ($/1M) for input and output separately. This is the industry-standard unit since 2023. Older per-1K-token pricing is converted to $/1M by multiplying by 1,000 for display consistency.

Input vs. output rate asymmetry

All current frontier models charge more for output tokens than input tokens — typically 3-5x more. Anthropic Claude 3.5 Sonnet: $3/1M input, $15/1M output. OpenAI GPT-4o: $2.50/1M input, $10/1M output. This asymmetry drives the cost of generation-heavy tasks (writing, summarization) versus classification or extraction tasks.

Our cost model computes: total_cost = (input_tokens × input_rate) + (output_tokens × output_rate). For tasks where the output/input ratio is known, we model it as a ratio parameter (e.g., 1:3 for typical chat completions, 1:8 for long-form generation).

Rate update protocol

LLM pricing changes frequently — Anthropic has repriced models 3 times in 2 years, and GPT-4 pricing has dropped by 97% from initial release. We update pricing when providers announce changes, flagging the date of last update on each calculator. Users should verify current pricing against provider pages for production cost modeling.

Limitations

Enterprise pricing is typically negotiated below list rates — our calculators use public list rates. API pricing excludes Anthropic Claude.ai subscription, ChatGPT Plus subscription, and other consumer products. Token counts are estimates — actual tokenization depends on the specific tokenizer (tiktoken for OpenAI, claude-tokenizer for Anthropic), which we approximate using the 4-characters-per-token heuristic for English text.

Update protocol

This category is reviewed quarterly. Immediate updates are triggered by changes to the primary source documents listed in the citations above — rate table revisions, new agency guidance, or regulatory amendments.

Error reports go to info@bedrockatools.com. Corrections are published on our corrections page.