LLM Economics

Cheapest LLM for Summarization

Compare AI model costs for document summarization. Find the most cost-effective model for processing long documents at scale.

Inputs

Select a use case to pre-fill realistic token and request defaults, or choose Custom.

The model you are using in production today.

Average prompt size per API call — system prompt + context + user message. Typical: simple chat 500–2,000 · RAG chatbot 2,000–8,000 · document analysis 10,000–100,000 tokens.

Average response length. Short answer: 200–500 · paragraph: 500–1,000 · long-form: 1,000–4,000 tokens.

Total API calls per month. 1,000 DAU × 5 calls/day × 30 days = 150,000 requests/month.

Fraction of input tokens currently served from Prompt Cache. 0 = not using cache.

Max potential savings
$2,748.78/year

Decision Summary

Best move
Best combo: GPT-5 Nano + cache + batch
Expected savings
$2,748.78 /yr (98.5% reduction)
Watch out
Validate quality, latency, and reliability on your own workload before changing production traffic.
Next step
Model migration ROI
Best combo: GPT-5 Nano + cache + batch
Best combo: GPT-5 Nano + cache + batch saves $2,748.78/yr with high implementation effort.
$2,748.78
saved / year (98.5%)
Top opportunity: Best combo: GPT-5 Nano + cache + batchHigh effort

Saves $2,748.78/yr (98.5%) with high engineering effort. See the ranked table below for all options.

$2,748.78
saved / year
Current yearly cost
$2,790.00
Current monthly cost
$232.50
Best optimised cost
$3.44/month
Top action
Best combo: GPT-5 Nano + cache + batch

Comparison

OptionMonthlyYearly
Best combo: GPT-5 Nano + cache + batchMaximum possible savings: optimal model, 60% cache, batch API enabled.cheapest$3.44$41.22
Switch to GPT-5 NanoRun a representative quality evaluation against the current model before migrating traffic.$6.87$82.44
Switch to GPT-5 MiniRun a representative quality evaluation against the current model before migrating traffic.$34.35$412.20
Cache (60%) + Batch APIMaximum savings without changing models.$85.88$1,030.50
Switch to Claude Haiku 4.5Run a representative quality evaluation against the current model before migrating traffic.$101.40$1,216.80
Enable Batch API (50% discount)Use async batch endpoint. Suitable for offline / non-realtime workloads.$116.25$1,395.00
Enable Prompt Cache (60% hit)Fix system prompt at context start and reuse across requests.$171.75$2,061.00
Current baselineGPT-5, cache 0%, batch offcurrent$232.50$2,790.00

Pricing sources

Last verified 2026-08-03 · developers.openai.com/api/docs/models/compare developers.openai.com/api/docs/models/compare · platform.claude.com/docs/en/about-claude/pricing platform.claude.com/docs/en/about-claude/pricing · ai.google.dev/gemini-api/docs/pricing ai.google.dev/gemini-api/docs/pricing · api-docs.deepseek.com/quick_start/pricing api-docs.deepseek.com/quick_start/pricing

Trends & comparison

Trend

Comparison (monthly vs. yearly)

Summarization cost structure

Summarization is input-dominant: 6,000 input tokens (the document) vs 800 output tokens (the summary). This means input pricing matters most. Models like DeepSeek V4 Flash with $0.14/MTok input dramatically undercut GPT-5's $1.25/MTok. At 15,000 documents/month, the input cost alone ranges from $1.26 (DeepSeek) to $112.50 (GPT-5).

Batch processing for summarization

Most summarization workloads are not real-time — documents are processed in pipelines, overnight jobs, or on-upload queues. This makes them ideal for Batch API (50% discount). Combined with a cheap model, you can summarize 15,000 documents for under $5/month.

Frequently asked questions

Which LLM is cheapest for summarization?

GPT-5 Nano at $0.05/$0.40 per million tokens and DeepSeek V4 Flash at $0.14/$0.28 are the cheapest options. Summarization is input-heavy, so models with low input pricing win. Gemini Flash-Lite is also competitive.

How much does AI summarization cost at scale?

At 15,000 documents/month with 6,000 input + 800 output tokens each: GPT-5 costs ~$232/month, GPT-5 Mini ~$47, GPT-5 Nano ~$9. Batch API saves another 50% for non-realtime pipelines.

Can cheap models summarize well?

Yes. Summarization is one of the most model-tolerant tasks. Even small models produce acceptable summaries for most content. The main risk is missing key details in long documents — validate with a sample before full deployment.

Related calculators

Compare verified first-party API prices on the same standard workload → AI Model Price Leaderboard

Cheapest LLM for Summarization — AI Document Processing Cost 2026