The best AI model for proofreading (we tested them all)

As of July 24 2026, GPT-5.6 Sol is the best AI model for proofreading, scoring 96.5% with medium reasoning, while missing just 2.3% of errors and making bad fixes on only 1.2%. But the new story is second place: Claude Opus 5 debuts at 95.7% with low reasoning - and it costs less than Sol ($0.96 vs. $1.27 per benchmark run) while fixing errors nearly three times as fast, at 22.5 fixes per minute to Sol’s 8.4.

So if you want the highest-quality result, use GPT-5.6 Sol with medium reasoning. Claude Opus 5 is the new pick when speed or cost matters - it gives up less than a point of quality and is the fastest model in the top tier - while GPT-5.6 Terra trades a few more points of quality for an even lower price.

Some other noteworthy mentions based on the full data:

• The smart-money pick: Grok 4.5. It scores 93.4% with high reasoning - 3.1 points behind Sol at less than half the cost. If you don’t need the absolute best result, this is still an excellent value.
• Gemini 3 Flash remains a great budget pass - it scores 92.5% at a similar price to Grok 4.5, though Grok edges it out on quality.
• Opus models like their reasoning dial low. Opus 5 scores best with low reasoning (its no-reasoning variant drops all the way to 85.6%), and the older Opus models (4.6, 4.7, 4.8) do best at proofreading with no reasoning at all!
• Kimi K3 is the newcomer to watch. Moonshot AI’s latest debuts in 7th place at 92.1% with max reasoning, tying Gemini 3.1 Pro - and its 0.6% bad-fix rate is tied for the lowest of any model in the top ten. The catch is speed: at about 3 fixes per minute it’s one of the slowest models we tested.
• GLM 5.1 and 5.2 are mention-worthy - they offer a cost-quality trade-off that no other model can beat, with GLM 5.1 scoring just under 90%.
• Grok 4.1 (medium) is an even more cost-effective option, if you're okay with a modest quality drop. It scored just over 85% on our benchmark and costs about 15x less than GPT-5.5.
• There are certain models it just never makes sense to use, such as Claude Sonnet (inferior to Gemini and GLM) and Grok 4.2 - beaten on both quality and price by xAI’s own 4.3 and 4.5.

The top 5 models by quality overall are below. See the full results for more data and scatterplots.

A proofreading benchmark only means something if the test is realistic. Here’s the short version of how ErrataBench works:

• Real text, real errors. We start from genuine source documents across 11 datasets - novels, legal text, scientific writing, nearly 99,000 words in all - and use a model pair to seed small, realistic mistakes drawn from a taxonomy of error categories, with a second model reviewing each one before it’s accepted.
• No cheat sheet. Each model is asked to proofread carefully, but is never told what kinds of errors are hidden or how many there are - just like a real editing job.
• The same tools for everyone. Models work through a simple agent loop with find-and-replace tools, get the same text chunks and the same number of turns, and are graded by an LLM judge on whether each fix is actually correct.

Every model gets the identical prompt, tools, and text - so the only variable is the model itself. Full methodology, source code, and raw results are published and reproducible.

Beyond the headline fix rate, two numbers tracked in the full results matter just as much: missed errors - the ones a model never touched (a completeness problem) - and bad fixes, the changes it attempted that were actually wrong. The second is the dangerous one - a model that “corrects” text incorrectly makes your writing worse. The best models keep bad fixes near 1%.

The most accurate model isn’t always the one you should reach for. Proofreading is often high-volume - a long report, a whole manuscript, a stack of documents - so cost and speed matter.

At the very top, GPT-5.6 Sol leads on quality at about $1.27 per benchmark run - but Claude Opus 5 is the smart-money flagship: 95.7% for about $0.96, running nearly three times faster than Sol and undercutting GPT-5.5 ($1.88) by nearly half. Below the flagships, Grok 4.5 reaches 93.4% for about $0.60, and GPT-5.6 Terra reaches 93.1% for about $0.72 while running faster than Sol. Gemini 3 Flash scores 92.5% at a similar price, while Gemini 3.1 Pro reaches 92.1% for roughly $0.37. Kimi K3 matches that 92.1% too, but at about $0.91 per run and a much slower pace, Gemini is the better buy at that quality level.

That’s the whole argument for not marrying a single model: use a fast, cheap one for quick passes and escalate to a flagship for the writing that really matters. You can explore the full accuracy-vs-cost tradeoff - plus speed and consistency - on the interactive ErrataBench scatterplot.

No model is perfect, and the errors they miss aren’t random. Aggregated across every model we tested, the easiest category to catch is spelling & word form - the obvious typos a basic checker would flag too, resolved about 91% of the time. The hardest is meaning, flow & style, resolved only around 79% of the time - the subtle, judgment-heavy issues where reasonable editors can disagree. The full results break performance down by error category.

The practical lesson: AI proofreading is excellent, but it’s a collaborator, not an oracle. You still want to see and approve what it changes - which is why the how of AI proofreading matters as much as which model.

Picking a model is half the battle. The other half is the tool you run it in. Pasting paragraphs into a chatbot and copying results back is slow, loses your formatting, and - worst of all - gives you no easy way to see exactly what changed.

Revise is a word processor with the AI built into the editor, designed for exactly this:

• Every top model, one click away. Because Revise is multi-provider, you can proofread with GPT-5.6 Sol, run a faster pass with Claude Opus 5, and get a cheaper second opinion from Gemini 3.1 Pro - all in the same document, no accounts or API keys to juggle.
• Whole-document context. The AI reads your entire draft, so it catches inconsistencies a paragraph-at-a-time chatbot never would - a term spelled two ways, a tense that drifts.
• Tracked changes you control. Every fix appears as a red/green diff you accept or reject one by one. The AI proposes; you decide. Nothing is overwritten silently.
• Full revision history. Scrub back through every version to see exactly how the document evolved and which edits came from you versus the AI.

In other words, Revise turns “which model is best” from a one-time bet into a dial you can turn per task - with the review controls that make it safe to let AI touch a document that matters.

What’s the best AI model for proofreading?

GPT-5.6 Sol with medium reasoning. It tops ErrataBench at 96.5%, less than a point ahead of Claude Opus 5’s 95.7%. Opus 5 is the better pick when speed or cost matters - it’s cheaper than Sol and nearly three times faster - while Grok 4.5 is the value pick at 93.4% for less than half Sol’s cost.

Is ChatGPT or Claude better at it?

ChatGPT still leads on accuracy, but barely: GPT-5.6 Sol beats the new Claude Opus 5 by 0.8 percentage points. Claude wins on everything else - Opus 5 costs less per run and fixes about 22.5 errors per minute versus Sol’s 8.4, and its 0.9% bad-fix rate is the lowest of the top flagships. The better choice depends on whether that last point of accuracy or turnaround time matters more.

How accurate is AI proofreading, really?

The leaders find and correctly fix more than 90% of errors (the best, GPT-5.6 Sol, hits 96.5%) and rarely break anything - but they’re strongest on mechanical mistakes and weakest on judgment calls involving meaning, flow, and style. Treat the AI as a collaborator and review its changes.

And since the models keep leaping ahead of each other, the smartest move isn’t to pick one forever - it’s to use a tool that lets you switch. Try it on your own writing: paste a document and proofread it with the top AI models. It’s free to use with basic AI, and with a paid upgrade it offers the most powerful AI features of any word processor. And if you want to go deeper on the data, the full ErrataBench results are open and interactive.

Quoting this post? All the data here is free to reuse with a link back. Copy a ready-made citation.