Gemma 4 31B

Apr 2026
Gemma 4 31B
Compare in
on
—per task?
$0.10 / $0.34per 1M in / out?
55 tok/sfirst token in 1.8 s?
256Kcontext

Rankingin thinking mode?

Strong at

S
Instructions#21 of 245
A
Science#71 of 312
B
Tool calling#102 of 242Reasoning#105 of 334Long context#126 of 273Vision#61 of 124
C
Coding#75 of 132Agents & work tasks#113 of 175Knowledge#198 of 318

Alternatives?

ModelScoreCostPer taskWhat changes
BetterGPT-6 Astraunstated+59%+12,400%$10 / $50Monitoring can return HTTP 403+2OpenMuse Glimmer 30Bunstated+31%+212%$0.30 / $1.10Context: 256K → 131KBetterClaude Fable 5.1unstated+58%+12,400%$10 / $50Replace forced tool_choice+5BetterClaude Opus 5unstated+56%+6,150%$5 / $25BetterGPT-6 Solunstated+55%+2,400%$2 / $10

Run it yourself

estimate
Fits on✗ 8 GB (phone, laptop)✗ 16 GB✓ 24 GB GPU · compressed✓ 48 GB GPU · 8-bit✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam23.6#105
AA Omniscience Index-47.9#199
CritPt (physics)1.4#113
AA Intelligence Index 4.3.214.7#148
AA Omniscience · Accuracy20#155
AA Omniscience · Hallucination Rate85#200
AA-LCR (long context) 1.169.7#126
GDP.pdf · AA all-pass6#78
GPQA Diamond85.7#71
Coding
Terminal-Bench 4.00#75
SciCode45.5#71
Terminal-Bench Hard (version unstated)36.4#39
SWE-bench Pro (edition unstated)lab35.7
Terminal-Bench 2.143.4#81
Agents & Tool Use
AutomationBench-AA4.9#85
GDPval-AA (work tasks) 2.16#113
τ³-Banking · AA14.8#71
ITBench SRE37.3#22
EnterpriseOps-Gym-AA28.3#38
AA-Briefcase 1.1 · Elo367.9#105
τ²-bench · AA59.9#104
Writing & Language
IFBench (instructions)75.6#21
Vision & Documents
MMMU-Pro (vision)73.4#61

FAQ

How much does Gemma 4 31B cost?›

$0.10 per 1M input tokens and $0.34 per 1M output tokens (cheapest host: CoreWeave).

Is Gemma 4 31B free to use?›

There is no free API; the cheapest paid price is $0.10 per 1M input tokens. The weights are open, so you can run it yourself.

How good is Gemma 4 31B, and what is it best at?›

#105 of 334 in Chat & Knowledge, #75 of 132 in Coding, #85 of 133 in Agents & Tool Use, #55 of 195 in Writing & Language, #61 of 124 in Vision & Documents. Strongest at: Science, Instructions.

Is Gemma 4 31B better than Mistral Large 3?›

On EQ-Bench Creative Writing v3, Gemma 4 31B scores 1,368.2 vs 1,412.2 — lower.

How can I use Gemma 4 31B?›

Through the API of Chutes, CoreWeave, Crusoe, DeepInfra, Featherless and 7 more.

What is the context window of Gemma 4 31B?›

256K tokens, with up to 33K output tokens.

When was Gemma 4 31B released? Is there a newer version?›

Released 2026-04-02. It is the current version in its line.

Is Gemma 4 31B open source?›

Yes, the weights are open (Apache-2.0 licence).

Can I run Gemma 4 31B locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can Gemma 4 31B read images?›

It accepts text and images, video as input.

What is Gemma 4 31B?›

Gemma 4 31B is an open-weights language model from Google, released Apr 2026.

What are the best alternatives to Gemma 4 31B?›

GPT-6 Astra (better), Muse Glimmer 30B (open weights).