Best AI for proofreading

As of 2026-07-18, Anthropic: Claude Opus 4.8 (Fast) is the strongest model for this task in our daily benchmark, where over 350 AI models are evaluated and the best are presented.

Proofreading is where AI delivers the most value at the least risk: the text is already yours, the model just has to find the errors. But it genuinely has to know the language — a model with weak grasp will “correct” correct Norwegian into incorrect Norwegian.

We rank by language quality, which is the direct measure of whether a model can correct the language at all. Ranked by language score, from today’s benchmark.

Runner-up
Anthropic: Claude Haiku 4.5
anthropic
Score: 8.0/10
Language: 9.3/10
Speed: 113 t/s
$1.00
Third choice
Google: Gemma 3 4B
google
Score: 8.0/10
Language: 9.3/10
Speed: 237 t/s
$0.05

Today’s ranking

Model Provider Score Language Speed Input price/1M
Anthropic: Claude Opus 4.8 (Fast) anthropic 8.3/10 9.3/10 131 t/s $10.00
Anthropic: Claude Haiku 4.5 anthropic 8.0/10 9.3/10 113 t/s $1.00
Google: Gemma 3 4B google 8.0/10 9.3/10 237 t/s $0.05
Cohere: Command R+ (08-2024) cohere 7.5/10 9.3/10 160 t/s $2.50
Perplexity: Sonar Pro perplexity 7.7/10 9.0/10 68 t/s $3.00
Meta: Llama 3.1 8B Instruct meta-llama 5.2/10 9.0/10 392 t/s $0.05
Mistral: Mistral Small 3 mistralai 7.8/10 8.7/10 234 t/s $0.05
IBM: Granite 4.0 Micro ibm-granite 5.4/10 8.3/10 179 t/s $0.02

Ask for correction, not rewriting

The default behaviour of most models is to rewrite the whole text in their own style. Be explicit: “Correct spelling, punctuation and grammar only. Keep the wording and sentence structure.” Asking it to list the changes shows you what was actually done.

What AI misses

Models are good at commas, agreement and compound words. They are weaker on factual errors, internal contradictions and whether the text actually works for the reader. Proofreading is not the same as editing.

hvilkenAI is independent. We have no affiliate agreements, sponsored placements or commercial ties to the AI providers we test. Every recommendation rests on test results alone.

Frequently asked questions

Which AI corrects Scandinavian languages best?
The ranking above comes from today’s benchmark, where we evaluate over 350 models and present the strongest on the language.
Can AI replace a proofreader?
For everyday text, largely yes. For text that will be published or carry legal weight, no — use AI as the first pass and a human as the last.
Does AI catch compound-word errors?
The strongest models catch most cases, but not all. Compound words are one of the error types where Scandinavian languages differ sharply from English and where weak models fail most often.
Will the AI change the meaning of my text?
It can, if you do not tell it not to. Explicitly ask it to preserve content and wording, and review the changes.

Other tasks

Doing something else? See which model is best for it.

→ Other tasks

Read also

← Back to home