Almost every AI model takes a million tokens now, but 12 of 40 charge more once a prompt gets long, and Claude Haiku 5.5 charges five times as much past 100K

37 of the 40 AI models on ArtificialWatch's price list accept a prompt of a million tokens or more; only xAI's three Groks stop at 500,000. What differs is the bill for using that room: 12 of the 40 charge a higher rate once a prompt passes a set length.

THE SURCHARGES

- Claude Haiku 5.5: five times the rate past 100,000 tokens, $0.10 to $0.50 per million input tokens and $0.50 to $2.50 output. It is the only one of Anthropic's eight models on the list with a long-prompt rate.

- OpenAI: all seven of its models on the list double the input rate and raise the output rate by half past 272,000 tokens, from GPT-6 Luna ($0.10 to $0.20 input) to GPT-6 Astra ($10 to $20).

- xAI: Grok 4.5, 4.6 and 4.7 double both rates past 200,000 tokens, inside their 500,000-token windows.

- MiniMax: MiniMax M3 doubles both rates past 512,000 tokens. MiniMax M3.1 Flash has one rate.

ONE RATE AT ANY LENGTH

The other 28 charge the same rate across the whole window: Google's six Gemini models on the list, seven of Anthropic's eight (Claude Opus 5.5, Claude Sonnet 5.5 and Claude Fable 5.1 among them), and every model on the list from Zhipu AI, Meta, Moonshot, Alibaba, Xiaomi, Mistral and DeepSeek. Gemini 4 Argon's rate does double on December 1, at any prompt length.

WHAT A LONG PROMPT COSTS

A 300,000-token prompt to GPT-6.1 Sol costs $1.20 in input at the long-prompt rate, against $0.60 at its standard rate. A 150,000-token prompt to Claude Haiku 5.5 costs $0.075 in input, against $0.015 at the rate below 100,000 tokens. Below the line Haiku 5.5 is cheaper than Gemini 3.5 Flash-Lite ($0.10 against $0.30 per million input tokens); above it, Haiku 5.5 costs more ($0.50 against $0.30).

THE WINDOWS

The 37 million-token windows run from 1,000,000 to 1,050,000 tokens, with OpenAI's the largest at 1,050,000. The window covers the prompt and the reply together.

HOW WE COUNTED

From ArtificialWatch's LLM price list as of November 4, 2026 (artificialwatch.com/llm-pricing, also at /data/llm-prices.json): each model maker's own API list price for the model's exact catalog id, with the long-prompt rate as the maker lists it. Clouds and resellers may charge differently. We checked Claude Haiku 5.5's rates on Anthropic's pricing page and OpenAI's 272,000-token line on OpenAI's. Google's pricing page does list a higher rate past 200,000 tokens for Gemini 3.1 Pro Preview and the deprecated Gemini 2.5 Pro, neither of which is on the list. Every model's window and long-prompt line is at artificialwatch.com/llm-context-windows.