Skip to content
VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Models741Deals30d268Policy30d67, 15 pendingSources1,014

Gemma 4 31B

GoogleReleased Apr 2, 2026Current

Score basis

Gemma 4 31B is capable in instruction following; and behind the leaders in multimodal tasks, long context, reasoning, and coding. Too few results yet to rate agentic tasks, safety, math, multilingual tasks, or factuality.

Capability profile

Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.

Capable1 capability
  1. Instruction Following−24.0%1 of 3
Limited4 capabilities
  1. Multimodal−25.6%5 of 6
  2. Long Context−26.9%2 of 3
  3. Reasoning−31.8%3 of 6
  4. Coding−39.3%8 of 10
Not rated5 capabilities

Too few results yet to rate Agentic, Safety, Math, Multilingual or Factuality.

Price

$0.17input$0.40outputper million tokens

From Artificial Analysis · 4 providers tracked · All prices

Cheaper than 79% of 331 priced models · 3:1 input-to-output blend, log scale

Evidence

111results on81benchmarks

  • 10 independently verified
  • 38 aggregator
  • 17 vendor-reported
  • 46 cross-referenced

From 11 sources · latest Oct 8, 2026 · How verification works

API features

Tool callingJSON modeReasoning

As listed by OpenRouter

Research

31 papers reference Gemma 4 31B

Gemma 4 31B benchmark results

111 results on 81 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.

Instruction Following

CapableFull instruction following ranking

24.0% behind the leader1 of 3 ranked benchmarks measured

Show 1 more instruction following resultHide 1 instruction following result

Multimodal

LimitedFull multimodal ranking

25.6% behind the leader5 of 6 ranked benchmarks measured

Show 6 more multimodal resultsHide 6 multimodal results

Long Context

LimitedFull long context ranking

26.9% behind the leader2 of 3 ranked benchmarks measured

Show 1 more long context resultHide 1 long context result

Reasoning

LimitedFull reasoning ranking

31.8% behind the leader3 of 6 ranked benchmarks measured

Show 8 more reasoning resultsHide 8 reasoning results

Coding

LimitedFull coding ranking

39.3% behind the leader8 of 10 ranked benchmarks measured

Show 4 more coding resultsHide 4 coding results

Agentic

Not enough dataFull agentic ranking

0 of 7 ranked benchmarks measured

Show 4 more agentic resultsHide 4 agentic results

Math

Not enough dataFull math ranking

0 of 5 ranked benchmarks measured

Show 1 more math resultHide 1 math result

Factuality

Not enough dataFull factuality ranking

0 of 4 ranked benchmarks measured

Show 3 more factuality resultsHide 3 factuality results

More results

Benchmarks outside the capability baskets. They are not ranked against a leader.

Show 49 more resultsHide 49 results

Gemma 4 31B: common questions

Who makes Gemma 4 31B?

Gemma 4 31B is made by Google.

When was Gemma 4 31B released?

Gemma 4 31B was released on Apr 2, 2026, according to Artificial Analysis.

What is Gemma 4 31B good at?

Gemma 4 31B is capable in instruction following; and behind the leaders in multimodal tasks, long context, reasoning, and coding. Too few results yet to rate agentic tasks, safety, math, multilingual tasks, or factuality.

How much does Gemma 4 31B cost?

Gemma 4 31B costs $0.17 per million input tokens and $0.40 per million output tokens, according to Artificial Analysis. We track its price at 4 providers. At a mix of three input tokens to one output token, it is cheaper than 79% of the 331 priced models we track.

How many benchmarks has Gemma 4 31B been tested on?

We track 111 results for Gemma 4 31B on 81 benchmarks from 11 sources, 10 of them independently verified. The latest was recorded on Oct 8, 2026.

Which API features does Gemma 4 31B support?

OpenRouter lists tool calling, json mode, and reasoning for Gemma 4 31B.

About this record

Where Gemma 4 31B's numbers come from, and every name it appears under.

Tracked since
Apr 30, 2026
Newest source mention
Sep 4, 2026

Where the results come from

Verification: 111 scores · 10 independently verified · 38 aggregator-attributed · 46 vendor cross-reference · 17 vendor-reported. How these tiers are assigned

From 11 sources on 7 sites. Hugging Face supplies 61 of them; the 10 independently verified results come from 5 sites. Bars are coloured by trust tier.

  • huggingface.co61
  • artificialanalysis.ai39
  • raw.githubusercontent.com4
  • api.llm-stats.com2
  • datasets-server.huggingface.co2
  • lmarena.ai2
  • epoch.ai1

Also known as

Gemma 4 31B (Non-reasoning)gemma 4 31b (reasoning)Gemma 4 31B ITgemma-4-31b-non-reasoninggemma4-31bgoogle.gemma-4-31bgoogle/gemma-4-31b-itgoogle/gemma-4-31bgoogle/gemma-4-31b-it-assistantgoogle/gemma-4-31b-it-qat-q4_0-unquantized-assistantgemma-4-31b-it:batchgemma-4-31b-it:freegemma-4-31b-it-ultragemma-4-31b-it-turbo