What is a context window?

The context window is the maximum number of tokens a model can consider at once, covering both your prompt and the generated completion. If your input plus expected output exceeds the window, the request fails or older content is truncated. Larger windows let you pass more documents, history or examples in a single call.

Are these prices and limits exact?

They are realistic published estimates at the time of the last update, but providers change pricing and limits frequently. Always confirm the current figures on the provider's official pricing and model pages before committing to production budgets.

Why do input and output tokens cost different amounts?

Generating output tokens is more compute-intensive than reading input tokens, so most providers charge several times more for output. This is why prompt engineering that reduces verbose completions can cut your bill significantly even when the input stays the same.

What does modality mean in this table?

Modality lists the input types a model accepts — text only, text plus images (vision), or text plus images and audio. Choosing a multimodal model lets you send screenshots, documents or audio in the same request, but those models are often pricier than text-only equivalents.

Does a bigger context window mean better answers?

Not automatically. A large window lets you supply more context, but models can still lose track of details buried in the middle of very long inputs. Use retrieval to pass only the most relevant chunks rather than dumping everything just because the window allows it.

What is the Model Context Length Reference Table?

A searchable, filterable reference table of major LLMs showing context window size, input and output cost per 1M tokens, supported modalities and training cutoff. Filter by provider, context size or modality in your browser. It runs free in your browser on Gera Tools, with nothing uploaded.

Model Context Length Reference Table

Name: Model Context Length Reference Table
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Picking the right LLM means balancing context window, price and modalities. This reference table puts the headline numbers for the major models in one searchable, filterable place so you can answer “which model fits my prompt and budget?” in seconds.

How it works

Each row lists a model’s provider, maximum context window (in tokens), input and output cost per one million tokens, supported modalities and training-data cutoff. Use the search box to jump to a model, the dropdowns to filter by provider or modality, and the column headers to sort by context size or cost. Everything runs locally in your browser.

What a context window actually limits

The context window is the maximum number of tokens the model can process in a single call — this includes your system prompt, conversation history, retrieved documents, tool definitions, and the expected length of the generated response. If the sum of inputs plus expected output exceeds the window, the request either fails outright or the model silently drops the oldest content to fit.

A rough rule of thumb: 1,000 tokens is about 750 English words, or about 3–4 pages of typical prose. A 128K context window can fit roughly 90,000–100,000 words — a full short novel. A 1M context window could hold several long novels simultaneously.

Why output tokens cost more than input tokens

Generating output tokens requires the model to do a full forward pass through the network for each token, one at a time. Reading input tokens is processed in parallel in a single pass. This asymmetry is why providers charge — often 3× to 5× more — for output tokens. The practical implication:

Prompts that produce long, verbose outputs cost significantly more than prompts with concise instructions.
Asking the model to explain its reasoning step-by-step adds tokens and cost. Chain-of-thought is powerful but not free.
Structured output (JSON with fixed keys) is often more token-efficient than prose answers.

Choosing a model for your use case

For long-document processing (legal contracts, research papers, codebases): prioritise context window first, then cost. A small context window forces repeated chunking and loses cross-document coherence.

For high-volume production APIs (classification, extraction, summarisation at scale): prioritise cost per token. The difference between a $0.15/M input model and a $3/M input model is 20× — which at 10 million calls per month means tens of thousands of dollars in monthly savings.

For multimodal tasks (screenshots, diagrams, scanned documents): filter to vision-capable models. Text-only models cannot accept images regardless of prompt engineering.

For coding and tool use: look at benchmark performance on coding tasks alongside context size, since complex codebases need both large windows and strong reasoning.

Context size is not a substitute for good retrieval

A 1M context window lets you stuff everything in — but models can still lose track of details buried in the middle of very long inputs, a phenomenon sometimes called the “lost in the middle” problem. For tasks where precision matters (answering questions about a specific clause in a 500-page contract), retrieval-augmented generation (RAG) that surfaces the most relevant sections tends to outperform brute-force context stuffing even when the window is large enough to fit everything.

Notes and caveats

Treat the figures as realistic estimates — providers revise pricing and limits often, so verify the current numbers on official pages before locking a budget. Context windows and pricing as of the last table update; check provider documentation for the most current numbers.