LLM API pricing comparison: every model, ranked

Section 1

Rankings by category

Four top-10 leaderboards covering the workloads that drive most API bills. Every number is computed from the pricing dataset at render time — retune the rates in lib/pricing.tsand this page updates on the next build.

Section 2

Full comparison matrix

Every tracked model, sortable by any column. Filter by provider, minimum context window, or maximum cost per request; search by model name or id. Costs are computed at render time fromlib/pricing.ts.

Section 3

Cost scenarios

Same workload, every model. The bar is proportional to the most expensive model in each scenario so you can eyeball the 10×-and-more spread without doing arithmetic.

Section 4

Every pair, one click away

300 head-to-head pages, generated programmatically from the pricing dataset. Every link resolves to a dedicated comparison page with side-by-side rates, scenario cost ladders, and volume projections.

Groupings reflect lib/pricing.ts providers: Anthropic · OpenAI · Google.

Section 5

Frequently asked questions

Every answer is composed from the current pricing dataset at render time — both the prose above and the FAQPage JSON-LD embedded for search engines.

Which LLM API is the cheapest?

Based on current pricing, GPT-5 nano is the cheapest model at $0.000250 per request for a standard workload of 1,000 input / 500 output tokens. If you care most about input-token rate at scale, GPT-5 nano offers the best per-token price at $0.05 per million input tokens.

How much does it cost to run an LLM chatbot?

A chatbot handling 10,000 messages per month (500 input / 200 output tokens each) costs between $1.05 on GPT-5 nano and $225.00 on Claude Opus 4.1. Use the comparison table above to pick the cheapest model that meets your quality bar.

Is GPT-5.4 cheaper than Claude Opus 4.7?

At a standard workload (1,000 input / 500 output tokens), GPT-5.4 costs $0.0100 per request while Claude Opus 4.7 costs $0.0201. GPT-5.4 is cheaper by $0.0101 per request — roughly 50% less.

What is the cheapest LLM for coding?

For coding tasks at 5,000 input / 2,000 output tokens, GPT-5 nano offers the best cost at $0.001050 per request. Budget-conscious teams can drop to Gemini 2.5 Flash-Lite at $0.001300 per request if quality permits — typically fine for boilerplate edits, riskier for architectural reasoning.

How do LLM API prices compare in 2026?

LLM API prices dropped roughly 80% from 2024 to 2026. Frontier models now cost $0.05–$15.00 per million input tokens, while budget models start at $0.05 per million. See the full comparison table above for every currently-tracked model.