Skip to content
ModelCap

Gemma 4 31B

Current board #26Arena #58Open weightsFree tier

Google·google/gemma-4-31b-it

Gemma 4 31B is a language model published by Google. 17 OpenRouter endpoints are published in the latest snapshot. It accepts image, text, video input and returns text. The published context limit is 262K tokens. The catalogue declares support for reasoning, tool use, structured outputs.

ModelCap Index · Current board
74.0#26
Measured · 66% support · score interval 74.0–83.4
Market Gravity
63.2
Secondary market signal
Input
$0.10
per 1M tokens
Output
$0.34
per 1M tokens
Context
262K
tokens
Providers
17
endpoints
Gateway spend
—
latest · Vercel share

Providers

Uptime measured over the last 30 minutes
ProviderInputOutputContextQuantUptime
DeepInfraturboCheapest$0.09$0.34262Kfp4100.0%
CoreWeavebf16$0.10$0.34262Kbf1698.92%
OpenInferencebf16$0.10$0.35262Kbf1699.92%
Venicebf16$0.12$0.36256Kbf16—
Chutesfp4$0.12$0.37131Kfp473.75%
DeepInfrafp8$0.13$0.38262Kfp884.03%
SiliconFlowfp8$0.13$0.40262Kfp869.43%
Novitabf16$0.14$0.40262Kbf1694.76%
Friendli$0.14$0.40262K—100.0%
Morphfp4$0.14$0.40175Kfp495.80%
Crusoe$0.14$0.40262K—98.76%
Parasailfp8$0.15$0.40262Kfp890.94%
Phala$0.15$0.46262K—97.82%
ModelRunfp4$0.22$0.55262Kfp4100.0%
Together$0.28$0.86262K—94.34%
SambaNova$0.38$1.15262K—99.81%
Cerebrasfp16$0.99$1.49131Kfp1697.88%

Hugging Face Inference Providers

Exact repository identity

Provider economics and telemetry published by the Hugging Face Router. These are operational signals, not benchmark results.

ProviderInputOutputContextTTFTThroughputFeatures
deepinfra
live
$0.13$0.38262K639 ms13.7 tok/stools · structured
novita
live
$0.14$0.40262K1116 ms6.0 tok/s—
together
live
$0.39$0.97262K2393 ms8.9 tok/stools · structured
cerebras
live
$0.99$1.49131K259 ms558.4 tok/stools · structured
featherless-ai
live
——————

Reasoning efforts

Best supported result can inform ModelCap

Published Max, X-High, High and other configurations for this model. These are sourced scores, not head-to-head essays.

EffortAgent scoreSource rankSampleSource
Default
Gemma 4 31B
-17.22%
CI -19.5–-15.0%
#45700K obs
56K sessions
LMArena Agent
Match 100%

Reported evaluation leads (12)

Excluded from score

Structured rows reported in this exact model repository at the pinned revision below. They are useful research leads, but remain quarantined until dataset revision, harness, configuration and identity pass ModelCap’s independent admission review.

Dataset / taskMetricValueRevisionProvenance
cais/hle
hle
—26.5
Unpinned
not published
Reported
Quarantined
Idavidrein/gpqa
diamond
—84.3
Unpinned
not published
Reported
Quarantined
LiquidAI/ifstruct-v1.0
ifstruct_v1
—95.9
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
mean
—62.4
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
text_content
—89.9
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
text_formatting
—69.3
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
layout
—57.4
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
chart
—15
Unpinned
not published
Reported
Quarantined
llamaindex/ParseBench
table
—80.6
Unpinned
not published
Reported
Quarantined
MMMU/MMMU_Pro
mmmu_pro_vision
—76.9
Unpinned
not published
Reported
Quarantined
sapbot/ask-my-agent-bench-2
fallback
—72.1
Unpinned
not published
Reported
Quarantined
TIGER-Lab/MMLU-Pro
mmlu_pro
—85.2
Unpinned
not published
Reported
Quarantined

ModelCap Index · Current board

Methodology
74.0
Measured · 3 public benchmark observations across 3 boards
Current board
#26
Public rank basis
Measured (measured)
Index support
66.4% · strong
Index interval
74.0–83.4
Rank posterior
Not available
Top 5 / Top 10 probability
Not available / Not available
Posterior as of
Snapshot timestamp unavailable
Identity binding
Exact catalogue product · google/google/gemma-4-31b-it · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata google/gemma-4-31B-it@842da3794eaa0b77d5f08bae87a17459d91ff475 (not the evaluation revision)
Direct benchmark observations
3
Breadth-adjusted BES point estimate
78.7
Bayesian Evidence Score · measured evidence input only

BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.

Observed capability
78.7
Evidence status
confirmed · 88% mass
General preference
83.2
Coding
82.8
Agents & tools
7.4
Reasoning
—
Evidence breadth
97%
Evidence coverage
66%
Market and catalogue signals · secondary, never the language Index rank
Usage (OpenRouter popularity)
42.1
Liquidity (providers × uptime)
91.0
Open reach (HF downloads)
94.9
Surface (context / tools / modalities)
90.3
Freshness
61.6
  • · OpenRouter popularity #27
  • · 16 live providers
  • · HF 30d downloads 11028273
  • · Catalogue: context 262144, reasoning, tools, modalities Image/Text/Video
  • · First seen 2026-04-02

These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.

BES evidence selected · measured input only

The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.

Deployment readiness

Metadata index
93.8
strong · 98% metadata coverage
Not capability
Artifact reproducibility
100.0
Access & legal clarity
100.0
Deployability
98.0
Evaluation provenance
72.0

Missing: evaluationProvenance.evaluation-dataset-revisions.

A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.

Market activity

Venue coverage
Market Gravity
63.2
OpenRouter weekly popularity
#27 of 298

Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.

Arena preference

Status
Ranked
Source
Arena
Official rank
#58 of 385
Rating
1450.9
95% confidence interval
1443–1459
Votes
6K
Source snapshot
6 Aug 2026, 13:00 UTC
Category
overall
Identity match confidence
100%

Market Gravity breakdown

Usage55%
42.1

OpenRouter weekly popularity across the full model catalogue.

Liquidity25%
91.0

Independent providers versus the model's open or closed cohort, weighted by uptime.

Open reach15%
94.9

Hugging Face 30-day downloads for open models; neutral for closed models.

Freshness5%
61.6

Time since first public availability, on a six-month half-life.

Specification

Context window
262K tokens
Max output
262K tokens
Inputs
Image, Text, Video
Outputs
Text
Tokenizer
Gemma
Cached input tokens
$0.10 / 1M
First seen on OpenRouter
2 Apr 2026

Capabilities

  • Supported: Reasoning
  • Supported: Tool use
  • Supported: Structured output
  • Supported: Response format
  • Not supported: Moderated

Weights & access

Open weights. Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI.

Access
Open weights
Licence
apache-2.0
Pinned revision
842da3794eaa
Parameters
31.3B
Architecture
Gemma4ForConditionalGeneration
Model type
gemma4
Hugging Face downloads (30d)
11.0M
Downloads (all time)
45.7M
Likes
3K
Repository
google/gemma-4-31B-it
Repository updated
20 Jul 2026, 16:42 UTC
Base lineage
google/gemma-4-31B