LLM pricing · Report

LLM pricing report, November 2026: what an AI chat subscriber costs

Token prices do not tell you whether a subscription pays. This report turns the list prices of 56 OpenAI, Anthropic and Google text models into the numbers that do: what one chat message costs (1,500 tokens in, 400 out) and what one user costs a month on a $9.99 plan. Updated 2026-11-08.

Key findings

• A message costs from $0.00023 to $0.465. GPT-5 nano is the cheapest and o1-pro the most expensive, a 1,979× spread for the same conversation.
• A heavy user loses you money on 39 of 56 models. At 3,000 messages a month, their AI cost passes the $8.49 you keep from a $9.99 plan. A typical user (300 messages) does on 10.
• Output is the expensive half. The median model charges 5.0× more per output token than per input token, so long replies cost more than long prompts.
• Caching cuts the median message by 24%. That is for the 49 models with a cached-input price, when two thirds of each prompt (system prompt and history) is read from the cache.
• Announced price changes: Gemini 4 Argon from 2026-12-01 ($0.0070 → $0.0140 a message); Gemini 3.8 Flash from 2027-01-01 ($0.0026 → $0.0053 a message); Gemini 3.6 Flash from 2027-01-01 ($0.0026 → $0.0053 a message).

What a heavy user costs

Most AI apps have a few users like this, and a flat plan charges them the same as everyone else. Choosing a cheaper model for them, or capping them, decides whether the plan makes money.

By provider

Ten cheapest models per message

1. GPT-5 nano (OpenAI): $0.00023
2. GPT-4.1 nano (OpenAI): $0.00031
3. Gemini 2.5 Flash Lite (Google): $0.00031
4. GPT-6 luna (OpenAI): $0.00035
5. Claude Haiku 5.5 (Anthropic): $0.00035
6. GPT-4o mini (OpenAI): $0.00047
7. GPT-5.6 luna (OpenAI): $0.00078
8. GPT-5.4 nano (OpenAI): $0.00080
9. Gemini 3.1 Flash Lite (Google): $0.00097
10. GPT-5 mini (OpenAI): $0.0012

All 56 models, with a calculator: LLM API pricing compared.

Method and sources

Standard-tier list prices for prompts under each model's long-context threshold, audited against OpenAI, Anthropic and Google pricing pages (ACCURACY.md). A message is 1,500 input and 400 output tokens. Users: light user 30 messages a month, typical user 300 messages a month, heavy user 3,000 messages a month. Revenue: a $9.99 monthly plan less a 15% app store fee. Batch, priority and regional prices are not included. The data is MIT licensed.

To cite: FiscaLLM, "LLM pricing report, November 2026: what an AI chat subscriber costs", https://fiscallm.com/llm-pricing/report

Your own users

These are averages. FiscaLLM prices every AI call your app makes, per user, and sets it next to what that user paid through RevenueCat, Stripe or a CSV, so you see which users and plans lose money. Free and open source. See the demo or start free.