Gemma 4 31B
Googlefamily · Gemma 4huggingface.co/google/gemma-4-31B-it
Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...
Updated 3 h ago · first seen 11 Sept 2026
model_01M294WW1MERCRX1PDYT3XGFR9
Overview
Identity
- Canonical model
- Yesidentity confidence: mediumOne row per real model release. Artifacts (checkpoints, quantisations, conversions) and folded evaluation variants point here.
- Official checkpoints
- official_checkpoints = hf_repo identifiers carried by the model itself; artifacts are separate entities pointing here through canonical_id.
- Artifacts
- 4 quantizations0 official · 4 third-partySeparate entities (checkpoint · quantization · conversion · packaging) pointing to this model through canonical_id.
- Provider deployments
- 3
- API aliases
- gemma-4-31bgemma-4-31b-non-reasoninggoogle/gemma-4-31b-itIdentifiers under which providers and evaluators refer to this model.
- Folded evaluation variants
- 2Effort / thinking variants (…-high, …-non-reasoning) are result configurations of this model, not separate models. Their old URLs redirect here.
Openness
Open weights— weights downloadable under Apache-2.0; commercial use allowed; redistribution allowed; derivatives allowed; 4 dimensions unknown.
Weights downloadable under a permissive or Creative Commons licence allowing commercial use; code or data may be missing.
Weights
Yes
Inference code
—
Training code
—
Training data
—
Dataset
—
Commercial use
Yes
Redistribution
Yes
Derivatives
Yes
Licence: Apache License 2.0 (permissive · SPDX Apache-2.0 · stated as “apache-2.0”)
dimensions marked null are unknown, not false
Key facts
- Release date
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 9 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Openrouter id
Source:OpenRouter public model & pricing listingT2observed 13 h agomedium
Architecture
- Architecture
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Model type
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Parameters
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Tokenizer
Source:OpenRouter public model & pricing listingT2observed 13 h agomedium
- Weights dtype
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- File size
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Library name
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Pipeline tag
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium
- Hugging Face repo
Source:OpenRouter public model & pricing listingT2observed 3 h agomedium
Capabilities
Modalities
- Modalities
- imagetextvideo
- Input
- imagetextvideo
- Output
- text
Capabilities
Tool calling
Yes
OpenRouter public model & pricing listing · T2
Structured output
Yes
OpenRouter public model & pricing listing · T2
Reasoning
Yes
Artificial Analysis · T2
Vision
Yes
OpenRouter public model & pricing listing · T2
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
Source:OpenRouter public model & pricing listingT2observed 10 h agomedium
- Max output
Source:OpenRouter public model & pricing listingT2observed 13 h agomedium
- Tokenizer
Source:OpenRouter public model & pricing listingT2observed 13 h agomedium
Benchmarks36
Compare with another model →Comparable same task and conditions · Partially comparable same task, conditions differ (effort, temperature, judge) · Not comparable different variant or metric
Current rows only, grouped by benchmark → canonical metric → comparability group (task configuration). Effort variants folded into this model appear as rows of the same group. 36 current rows in total. “vs leader” compares with the current leader of the benchmark's primary group only; other groups are not directly comparable. Comparability rules →
Providers & Pricing7
All offers in the price terminal →USD per 1M tokens as published by each provider; native units (per-request fees, flex/priority tiers) are kept verbatim. Rows are append-only — every price change is kept in the history below. Cost of a workload →
Price history
Output price · USD / 1M tokens 3 providers
- Google Gemini API
- Together AI
- OpenRouter
- OpenRouter$0.97 → $012 Sept 2026
- OpenRouter$0.34 → $0.9712 Sept 2026
- OpenRouterfirst observed $0.3412 Sept 2026
- Together AIfirst observed $0.9711 Sept 2026
- Google Gemini API$0.97 → $011 Sept 2026
- Google Gemini API$0.34 → $0.9711 Sept 2026
- Google Gemini APIfirst observed $0.3411 Sept 2026
Input price · USD / 1M tokens 3 providers
- Google Gemini API
- Together AI
- OpenRouter
- OpenRouter$0.39 → $012 Sept 2026
- OpenRouter$0.09 → $0.3912 Sept 2026
- OpenRouterfirst observed $0.0912 Sept 2026
- Together AIfirst observed $0.3911 Sept 2026
- Google Gemini API$0.39 → $011 Sept 2026
- Google Gemini API$0.09 → $0.3911 Sept 2026
- Google Gemini APIfirst observed $0.0911 Sept 2026
Hardware fit37
Assumptions (7)
- Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
- bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
- KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
- A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
- Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
- Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
- Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.
Lineage
Open in Graph →Versions & Artifacts4
Version history
Context window2 changes
11 Sept 2026→11 Sept 2026→12 Sept 2026current
Licensefirst observation only
11 Sept 2026current
Max outputfirst observation only
11 Sept 2026current
Opennessfirst observation only
11 Sept 2026current
Parametersfirst observation only
11 Sept 2026current
Each hop is a claim: click a value for its source, tier and observation time. Nothing is overwritten — a new observation closes the previous claim.
Artifacts 4
quantizations 4
- Intel/gemma-4-31B-it-int4-AutoRoundIntel · BF16/F16/I3219.2 GB
- mlx-community/gemma-4-31b-it-4bitMLX Community · MLX · BF16/U3218.4 GB
- nvidia/Gemma-4-31B-IT-NVFP4NVIDIA · BF16/U832.6 GB
- unsloth/gemma-4-31B-it-qat-GGUFUnsloth · GGUF—
Papers1
- arXiv:2607.02770Active35
Timeline26
Full timeline →OpenRouter lists Gemma 4 31B at $0 in / $0 out per 1M tokens
openrouterOpenRouter lists Gemma 4 31B at $0.39 in / $0.97 out per 1M tokens
openrouterOpenRouter lists Gemma 4 31B at $0.09 in / $0.34 out per 1M tokens
openrouterGemma 4 31B: hf repo changed from google/gemma-4-31B to google/gemma-4-31B-it
Hugging Face repogoogle/gemma-4-31B→google/gemma-4-31B-itopenrouterGemma 4 31B scores 36.36% on Terminal-Bench
artificial_analysisGemma 4 31B scores 43.45% on Terminal-Bench
artificial_analysisGemma 4 31B scores 23.63% on Humanity's Last Exam
artificial_analysisGemma 4 31B scores 15.42 on Artificial Analysis Intelligence Index
artificial_analysisGemma 4 31B scores 29.21% on Terminal-Bench
artificial_analysisGemma 4 31B scores 11.82% on Humanity's Last Exam
artificial_analysisGemma 4 31B scores 13.92 on Artificial Analysis Intelligence Index
artificial_analysisGemma 4 31B: reasoning changed from false to true
ReasoningNo→Yesartificial_analysisGemma 4 31B: context length changed from 256000 to 262144
Context window256K tokens→262.1K tokensopenrouterGemma 4 31B: release date changed from 2026-03-11 to 2026-04-02
Release date11 Mar 2026→2 Apr 2026openrouterGemma 4 31B: release date changed from 2026-04-02 to 2026-03-11
Release date2 Apr 2026→11 Mar 2026huggingface
Change history58
Viewing AI Atlas as of 12 Aug 2026 — attributes exactly as the atlas knew them on that day; later corrections are not shown.
Back to today →Gemma 4 31B was not yet in AI Atlas on 12 Aug 2026
Release daterelease_date6
Opennessopenness2
Licenselicense1
Architecturearchitecture1
Parametersparameter_count1
Context windowcontext_length4
Max outputmax_output_tokens1
Modalitiesmodalities1
Input modalitiesmodalities_input1
Output modalitiesmodalities_output1
Tokenizertokenizer1
Base modelbase_model1
File sizefile_size_gb1
Hugging Face repohf_repo3
Pipeline tagpipeline_tag1
Model cardmodel_card_url1
Downloadsmetric.downloads1
Likesmetric.likes3
Aa context windowaa_context_window1
Aa opennessaa_openness1
Commercial use allowedcommercial_use_allowed1
Derivatives allowedderivatives_allowed1
Descriptiondescription1
Gatedgated1
Hf inference providershf_inference_providers1
Last modifiedlast_modified1
Library namelibrary_name1
License urllicense_url1
Aa median output tokens per secondmetric.aa_median_output_tokens_per_second2
Downloads all timemetric.downloads_all_time1
Model typemodel_type1
Openrouter idopenrouter_id1
Openrouter listed atopenrouter_listed_at1
Reasoningreasoning3
Redistribution allowedredistribution_allowed1
Structured outputstructured_output1
Supported parameterssupported_parameters1
Tool callingtool_calling1
Visionvision1
Weights availableweights_available1
Weights dtypeweights_dtype1
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
Provenance
Attributed facts
44
Source tiers
T244
Freshest observation
3 h ago
Conflicts
None
Source documents 9
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.
Data quality (72/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →