PodcastIntelligence Snacks · Weekly conversations about AI, software and business, from the team behind Normie Mode.Listen →
Models · Moonshot AI

Kimi K3

Moonshot AI's Kimi model, released 17 Jul 2026.

Top three for

2 tasks, including studying and homework, maths and science.

Evidence

Independently tested. 24 independent results so far.

Cost

$17 per 1,000 typical requests, if you use it directly.

Reads at once

1,573 pages of about 500 words.

Where you can use it

The apps and plans that include it, cheapest first.

App and planPriceAccess
Kimi FreeFree
Called Adagio
Included, the default
Kimi Moderato$19/monthIncluded, the default

Where it ranks, task by task

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

TaskPositionTask score
Best AI for studying and homework3rd of 1282
Best AI for maths and science3rd of 1281
Best AI for planning and personal admin4th of 1471
Best AI for business4th of 856
Best AI for health questions4th of 1474
Best AI for legal questions4th of 870
Best AI for coding and building apps4th of 1072
Best AI for research4th of 1164
Best AI for accounting and finance4th of 856
Best AI for CVs, job applications and interviews4th of 1456
Best AI for presentations and PowerPoint4th of 852
Best AI for writing5th of 1453
Best AI for translation and languages6th of 1466
Best AI for documents, PDFs and notes6th of 1366
Best AI for Excel and spreadsheets7th of 860

Test results

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

TestResultPosition
People's votes on legal questions
LMArena
rating 15063rd of 14
People's votes on business and finance questions
LMArena
rating 14714th of 14
People's votes on back-and-forth conversations
LMArena
rating 14884th of 14
People's votes on expert questions
LMArena
rating 15204th of 14
People's votes on following instructions
LMArena
rating 14814th of 14
People's votes on long, detailed requests
LMArena
rating 14944th of 14
People's votes on science questions
LMArena
rating 15054th of 14
People's votes on maths questions
LMArena
rating 14934th of 13
People's votes on health questions
LMArena
rating 14864th of 13
People's votes on coding questions
LMArena
rating 15095th of 14
People's votes on writing and language
LMArena
rating 14695th of 14
People's votes on creative writing
LMArena
rating 14526th of 14
People's votes on questions in other languages
LMArena
rating 14666th of 14
People's votes overall
LMArena
rating 14766th of 14
Would a real maintainer accept its code?
Cognition
44%4th of 9
Research-level maths
Epoch AI
72%5th of 11
Building features in real software projects
Datacurve
69%4th of 8
Graduate-level science questions
NYU and others
93%6th of 11
Competition maths
Epoch AI
97%7th of 12
Professional tasks in banking, consulting and law
Mercor
51%5th of 8
Pulling information out of documents into tables
DTBench authors
91%9th of 13
Answering short factual questions correctly
Google DeepMind
51%8th of 11

Details

Maker
Moonshot AI
Released
17 Jul 2026
Status
Current
Replaces
Kimi K2.5
Name checked
Confirmed with Moonshot AI
Intelligence Snacks newsletter

The big AI ideas each week, in your inbox