Google

Gemini 3.8 Flash

Made by Google DeepMind. Released 2026-09-02. Accessibility not recorded. Cheapest standard rate we hold: $3.75 per million output tokens, no context split.

The seller has announced a change to $3.75 from 2027-01-01. That is published in advance, not in force.

What it costs

Tier	Kind	In	Out	Cached in	From
batch	text	$0.38	$1.88	—	seller
flex	text	$0.38	$1.88	—	seller
priority	text	$1.35	$6.75	—	seller
standard	text	$0.75	$3.75	—	seller

per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.

What it scores, by house

Benchmark	House	Score	Evaluated	Scaffold
chess puzzles	Epoch AI	0.6±0.05	2026-10-13	high
frontiermath erdos	Epoch AI	0.0±0.00	2026-09-02	high
frontiermath tier 4 v2	Epoch AI	0.2±0.07	2026-09-02	high
frontiermath tiers 1 3 v2	Epoch AI	0.7±0.03	2026-09-02	high
furniture assembly	Epoch AI	0.3±0.06	2026-09-10	high
gpqa diamond	Epoch AI	1.0±0.01	2026-09-02	high
mirrorcode	Epoch AI	0.4±0.12	2026-09-02	high
mystery game puzzles	Epoch AI	0.5±0.05	2026-09-02	high
otis mock aime 2024 2025near its ceiling	Epoch AI	1.0±0.01	2026-09-02	high
simpleqa verified	Epoch AI	0.7±0.01	2026-09-02	high
ale bench	collected by Epoch AI	1270	—	high
apex agents	collected by Epoch AI	0.7	—	unknown
arc aginear its ceiling	collected by Epoch AI	1.0	—	medium
arc aginear its ceiling	collected by Epoch AI	1.0	—	high
arc aginear its ceiling	collected by Epoch AI	0.9	—	low
arc agi 2	collected by Epoch AI	0.9	—	high
arc agi 2	collected by Epoch AI	0.8	—	medium
arc agi 2	collected by Epoch AI	0.8	—	low
blueprint bench 2	collected by Epoch AI	0.4	—	unknown
critpt	collected by Epoch AI	0.0	—	low
critpt	collected by Epoch AI	0.1	—	medium
critpt	collected by Epoch AI	0.2	—	high
cursorbench	collected by Epoch AI	0.4	—	high
cursorbench	collected by Epoch AI	0.4	—	medium
deepswe	collected by Epoch AI	0.7	—	mini-swe-agent
dtbenchnear its ceiling	collected by Epoch AI	0.9	—	medium
dtbenchnear its ceiling	collected by Epoch AI	1.0	—	unknown
dtbenchnear its ceiling	collected by Epoch AI	0.9	—	low
frontiercode	collected by Epoch AI	0.4	—	devin-cli
frontierswe	collected by Epoch AI	0.2	—	proximus
gdp pdf	collected by Epoch AI	0.2	—	medium
gdp pdf	collected by Epoch AI	0.2	—	high
geobench	collected by Epoch AI	4307	—	unknown
hle	collected by Epoch AI	0.4	—	unknown
lmca	collected by Epoch AI	52.9	—	medium
lmca	collected by Epoch AI	51.5	—	unknown
lmca	collected by Epoch AI	51.2	—	low
proofbench	collected by Epoch AI	0.5	—	unknown
rli	collected by Epoch AI	0.1	—	unknown
scicode	collected by Epoch AI	0.6	—	medium
scicode	collected by Epoch AI	0.6	—	high
scicode	collected by Epoch AI	0.5	—	low
simplebench	collected by Epoch AI	0.8	—	unknown
surface evolver bench	collected by Epoch AI	0.8	—	high
vending bench 2	collected by Epoch AI	5094	—	unknown
webdev arena	collected by Epoch AI	1568	—	high
weirdml	collected by Epoch AI	0.8	—	high
weirdml v3	collected by Epoch AI	0.1	—	gemini_cli

These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.

Every rate we hold · Why there is no single best model · Everything from Google