Claude vs ChatGPT: price, context and limits compared

Claude and ChatGPT are answers to the same brief from two different companies — Anthropic and OpenAI — and this page lines up what each vendor publishes about its flagship model: how it’s priced, how much it can read at once, and how much it can write back. It does not compare how good the answers are. That’s a different question, and one this page deliberately leaves alone.

What you can actually compare

Both vendors charge per unit of text processed, split into a rate for what you send in and a rate for what the model sends back. In both cases the output rate is set higher than the input rate — that’s standard across this whole category, not something specific to either company. The table above carries the current figures; treat it, not this prose, as the source.

Context window is the other axis worth understanding: it’s the amount of text — instructions, documents, conversation history — a model can hold in view for a single reply. Every model has a ceiling here, and the two flagships compared on this page sit at different points on that scale, detailed in the table. Related to context is the output limit: a separate cap on how much text the model can generate in one go, which matters for long documents or large batches of generated content rather than for everyday chat.

When the difference matters

For casual, everyday use, pricing structure rarely bites — a handful of questions costs a fraction of a cent either way. It starts to matter at volume: a business running thousands of requests a day feels a gap between input and output pricing quickly, and small differences in rate compound into real budget lines.

Context window matters when you’re feeding a model something long — a contract, a codebase, a full research paper — and need the whole thing considered at once rather than chopped into pieces. If your work involves short, self-contained questions, the ceiling on context rarely comes into play at all.

Output limits matter for the opposite reason: generating something long in a single pass, like a full report or a large block of code, rather than asking a model to summarise or answer briefly. Check the table for how the two compare before committing a workflow to one or the other.

What this page won’t tell you

This page compares published price and capacity, not quality or capability. It won’t tell you which model reasons better, writes more naturally, or makes fewer mistakes — those rankings are genuinely contested, shift with every model update, and depend heavily on the specific task you’re testing. If that’s what you’re after, independent public evaluation leaderboards track exactly that, and they update far more often than a page like this could responsibly claim to.

One disclosure worth being upfront about: Around itself is built using Claude. That’s precisely why this page sticks to numbers a vendor publishes and refuses to rank one assistant above another — we have a stake in one of the two, so we’ve deliberately left ourselves no room to tip the scale.

For the current figures side by side, including other models beyond these two flagships, use the interactive comparison tool.