Gemini 3.8 Flash

Sep 2026
Gemini 3.8 Flash
Compare in
on
$0.93per task?
$0.75 / $3.75per 1M in / out?
—output speed?
1Mcontext

Rankingat medium effort?

Full ranking →
1GPT-6 Astra2,173.32Fable 5.12,1623Opus 52,132.64GPT-6 Sol2,124.75Kimi K32,082.3
···
27GLM 5.21,756.528Gemini 3.8 Flash1,747.929Kimi K2.61,724.5

Strong at

S
Science#8 of 312Long context#7 of 273Knowledge#14 of 318Vision#5 of 124
A
Coding#30 of 132Agents & work tasks#37 of 175Reasoning#37 of 334

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperGPT-5.6 Lunaunstated+5%-70%$0.20 / $1.20BetterGPT-6 Astraunstated+24%+1,233%$10 / $50Monitoring can return HTTP 403+2OpenKimi K3unstated+19%+300%$3 / $15Sampling parameters are fixed+3CheaperMuse Glimmer 30BunstatedOpen+3%-67%$0.30 / $1.10Context: 1M → 131KBetterClaude Fable 5.1unstated+24%+1,233%$10 / $50Replace forced tool_choice+5

History

7 versions
1,2001,4001,6001,800Jul 25Jan 26Jul 26Release dateEQ-Bench Creative v32.0 · Feb3.63.73.8
1,2001,4001,6001,800Jul 25Jan 26Jul 26Release dateEQ-Bench Creative v32.0 · Feb3.63.8

Benchmarks

in Writing & Language
Chat & Knowledge
Humanity's Last Exam42.1#37
AA Omniscience Index28.6#14
CritPt (physics)12.3#50
AA Intelligence Index 4.3.239.8#31
AA Omniscience · Accuracy53#14
AA Omniscience · Hallucination Rate51.9#95
AA-LCR (long context) 1.184#7
GDP.pdf · AA all-pass22.8#15
ARC-AGI-1 · public evaluation97.8#12
ARC-AGI-1 · semi-private97.5#7
ARC-AGI-2 · public evaluation87.5#15
ARC-AGI-2 · semi-private82.9#15
ARC-AGI-3 · semi-private4#7
GPQA Diamond93.5#8
Coding
Terminal-Bench 4.019.7#30
DeepSWE 1.171#6
SciCode55.1#26
Terminal-Bench 2.183.9#15
Agents & Tool Use
AutomationBench-AA60.9#16
GDPval-AA (work tasks) 2.146.1#37
τ³-Banking · AA45.8#10
AA-Briefcase 1.1 · Elo1,181.5#43
Vision & Documents
MMMU-Pro (vision)84.2#5

Watch out 2

Sampling parameters are ignored›

temperature, top_p and top_k are ignored now and will cause HTTP 400 in future generations; remove them from requests. Source ↗

No minimal thinking level›

The minimal thinking level returns an error; avoid minimal. Source ↗

FAQ

How much does Gemini 3.8 Flash cost?›

$0.75 per 1M input tokens and $3.75 per 1M output tokens. A typical benchmark task costs about $0.93 at medium effort.

Is Gemini 3.8 Flash free to use?›

There is no free API; the cheapest paid price is $0.75 per 1M input tokens.

How good is Gemini 3.8 Flash, and what is it best at?›

#37 of 334 in Chat & Knowledge, #30 of 132 in Coding, #16 of 133 in Agents & Tool Use, #6 of 195 in Writing & Language, #5 of 124 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Reasoning, Science, Long context, Knowledge, Vision.

Is Gemini 3.8 Flash better than Gemini 3.7 Flash?›

On EQ-Bench Creative Writing v3, Gemini 3.8 Flash scores 1,747.9 vs 1,723.2 — higher.

How can I use Gemini 3.8 Flash?›

Through the Google API.

What is the context window of Gemini 3.8 Flash?›

1M tokens, with up to 66K output tokens.

When was Gemini 3.8 Flash released? Is there a newer version?›

Released 2026-09-02. It is the current version in its line.

Is Gemini 3.8 Flash open source?›

No, Gemini 3.8 Flash is a closed model available through an API. The closest open-weights option here is Kimi K3.

Can Gemini 3.8 Flash read images?›

It accepts text and images, PDF files, audio, video as input.

What is Gemini 3.8 Flash?›

Gemini 3.8 Flash is a closed language model from Google, released Sep 2026, the successor to Gemini 3.7 Flash.

What are the best alternatives to Gemini 3.8 Flash?›

GPT-5.6 Luna (cheaper), GPT-6 Astra (better), Kimi K3 (open weights).