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

Gemini 3.8 Flash

Google DeepMind's Gemini model, released 2 Sept 2026.

Top three for

13 tasks, including studying and homework, planning and personal admin, business.

Evidence

Independently tested. 21 independent results so far.

Cost

$4.13 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
Google AI Pro$19.99/monthIncluded, the default
Google AI Ultra$99.99/month
From $99.99 a month; a $199.99 tier has higher limits
Included, 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 homework1st of 1293
Best AI for planning and personal admin2nd of 1485
Best AI for business2nd of 881
Best AI for maths and science2nd of 1289
Best AI for legal questions2nd of 887
Best AI for documents, PDFs and notes2nd of 1381
Best AI for accounting and finance2nd of 881
Best AI for presentations and PowerPoint2nd of 873
Best AI for translation and languages3rd of 1481
Best AI for writing3rd of 1467
Best AI for health questions3rd of 1489
Best AI for CVs, job applications and interviews3rd of 1466
Best AI for Excel and spreadsheets3rd of 893
Best AI for coding and building apps5th of 1071
Best AI for research6th of 1155

Test results

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

TestResultPosition
People's votes on maths questions
LMArena
rating 15281st of 13
People's votes on back-and-forth conversations
LMArena
rating 15032nd of 14
People's votes on expert questions
LMArena
rating 15292nd of 14
People's votes on health questions
LMArena
rating 15082nd of 13
Graduate-level science questions
NYU and others
95%2nd of 11
People's votes on business and finance questions
LMArena
rating 14803rd of 14
People's votes on coding questions
LMArena
rating 15123rd of 14
People's votes on creative writing
LMArena
rating 14853rd of 14
People's votes on following instructions
LMArena
rating 14893rd of 14
People's votes on long, detailed requests
LMArena
rating 15063rd of 14
People's votes on questions in other languages
LMArena
rating 14863rd of 14
People's votes overall
LMArena
rating 14943rd of 14
People's votes on science questions
LMArena
rating 15223rd of 14
People's votes on writing and language
LMArena
rating 14893rd of 14
Building features in real software projects
Datacurve
74%2nd of 8
People's votes on legal questions
LMArena
rating 15004th of 14
Competition maths
Epoch AI
99%4th of 12
Answering short factual questions correctly
Google DeepMind
70%4th of 11
Professional tasks in banking, consulting and law
Mercor
64%3rd of 8
Pulling information out of documents into tables
DTBench authors
96%5th of 13
Answering business questions from a spreadsheetTested by us
Normie Mode
30/303rd of 6
Research-level maths
Epoch AI
68%6th of 11
Would a real maintainer accept its code?
Cognition
41%6th of 9
Humanity's Last Exam
Center for AI Safety, Scale AI
45%4th of 4

Details

Maker
Google DeepMind
Released
2 Sept 2026
Status
Current
Replaces
Gemini 3.6 Flash, Gemini 3.7 Flash
Name checked
Confirmed with Google
Intelligence Snacks newsletter

The big AI ideas each week, in your inbox