Checked against the OpenRouter public models feed. Prices are copied, not estimated.

The Best Model ForWhich AI model to use, from ten labs.

The best AI model for long documents

A contract, a transcript, or a pack of files that has to fit in one request.

Short answer

Start with Gemini 4 Hydrogen ($1.50 in · $7.50 out per million tokens). Hydrogen lists about a million tokens and accepts files. A 300-page contract is the motivating case. Ask for parties, dates, renewal, and unusual clauses. A cheaper stop is Llama 4 Scout ($0.10 in · $0.30 out).

Quality

Gemini 4 Hydrogen

Google (Gemini) · $1.50 in · $7.50 out

Hydrogen lists about a million tokens and accepts files. A 300-page contract is the motivating case. Ask for parties, dates, renewal, and unusual clauses.

Gemini 4 Hydrogen is listed at $1.50 input and $7.50 output per million tokens, with a 1,048,576-token context window.

Balanced

Grok 4.20

xAI (Grok) · $1.25 in · $2.50 out

The longest reviewed window on the feed, 2,000,000 tokens, with a cutoff of 2025-09-01. Use it when the pack of files is the product.

Grok 4.20 is listed at $1.25 input and $2.50 output per million tokens, with a 2,000,000-token context window.

Cheapest sensible

Llama 4 Scout

Meta (Llama) · $0.10 in · $0.30 out

Open weights, a window above a million tokens, image input, and a low rate. The listed cutoff is 2024-08-31.

Llama 4 Scout is listed at $0.10 input and $0.30 output per million tokens, with a 1,310,720-token context window.

Also on the feed

These match the factual limits of the task (modality, tools, context) and are not the reviewed picks. A row marked new has no guide note.

Questions

Does a million tokens equal 300 pages?

Often, not always. Scans, tables, and images eat the window faster than plain prose. If the request fails, the document did not fit, whatever a rule of thumb said.

Is the model a lawyer if it reads the contract?

No. It is a reader. Someone accountable still decides.