Models · DeepSeek

DeepSeek V4.1 Flash

DeepSeek's DeepSeek model, released 10 Sept 2026.

Top three for

No task yet.

Evidence

Independently tested. 6 independent results so far.

Cost

$0.72 per 1,000 typical requests, if you use it directly.

Reads at once

1,500 pages of about 500 words.

Where it ranks, task by task

Its place on each task's score, among the models we compare that have enough results to score.

Task	Position	Task score
Best AI for planning and personal admin	7th of 14	46
Best AI for translation and languages	7th of 14	56
Best AI for health questions	7th of 14	57
Best AI for writing	8th of 14	44
Best AI for CVs, job applications and interviews	8th of 14	46
Best AI for documents, PDFs and notes	10th of 13	55

Every test we use that has a result for this exact version.

Test	Result	Position
People's votes on coding questions LMArena	rating 1511	4th of 14
People's votes on maths questions LMArena	rating 1493	5th of 13
People's votes on health questions LMArena	rating 1475	5th of 13
People's votes on business and finance questions LMArena	rating 1464	6th of 14
People's votes on expert questions LMArena	rating 1505	6th of 14
People's votes on following instructions LMArena	rating 1475	6th of 14
People's votes on back-and-forth conversations LMArena	rating 1460	7th of 14
People's votes on long, detailed requests LMArena	rating 1476	7th of 14
People's votes on questions in other languages LMArena	rating 1452	7th of 14
People's votes overall LMArena	rating 1465	7th of 14
People's votes on science questions LMArena	rating 1471	7th of 14
People's votes on writing and language LMArena	rating 1448	8th of 14
People's votes on creative writing LMArena	rating 1436	9th of 14
Pulling information out of documents into tables DTBench authors	90%	10th of 13
People's votes on legal questions LMArena	rating 1439	12th of 14