Best AI for writing

Writing is where models diverge most on taste. Two models can score identically on instruction-following and produce prose with completely different voice — one tight and readable, the other syrupy and evasive. Benchmark scores alone won't tell you which model matches your brand.

What matters: rhetorical range, willingness to cut, instruction adherence at the paragraph level, and refusal discipline. Good models don't moralize at the drop of a hat. We weight side-by-side human preference strongly here, along with tone consistency across long documents.

Our writing rank weights writing quality (2.5×), instruction-following (1.5×), and reasoning (0.5×). For editorial work at scale, instruction-following is the main differentiator — models that silently ignore style constraints waste editorial time.

The best AI for writing changes every month.

We'll email you when rankings shift, new models hit the top 5, or pricing cuts reshuffle the value leaders.