Gemma 4 31B IT

Google
Model overview

gemma-4-31B-it

Google's current public Gemma 4 31B instruction-tuned open-weight model.

general LLMreasoningvision-languagechatlocal inference
Representative version

gemma-4-31B-it

CompanyGoogle
ReleaseNot disclosed
Release typeModel release
Parameters31B
ArchitectureNot disclosed
ContextNot disclosed
Input modalitiestext / image
Output modalitiestext
Tool use / function callingNot disclosed
Structured output / reasoningNot disclosed
Vision / audio / videoConfirmed: Vision; Not shown: Audio, Video
Open weights / LicenseWeights available / Apache-2.0
License sourcegoogle-gemma-4-31b-it-current · 2026-07-15
Vendor APINot disclosed
API source pending verification
OpenRouterOpenRouter listed
PricingOpenRouter · 2026-07-15: $0.22 / $0.55
input / output, per 1M tokens
BenchmarkNo reliable public source
Public signals3 runtime / access records / 1 activity snapshot
Use casesassistant chat / local inference / general text tasks / complex reasoning
Claim evidenceRelease / model specs / weightsgoogle/gemma-4-31B-it · Hugging Facegemma-4-31B-it
OpenRouter

Pricing / context / availability

OpenRouterOpenRouter listed
OpenRouter listinggoogle/gemma-4-31b-it
Mapped versiongemma-4-31B-it
google-gemma-4-31b-it-current
PricingOpenRouter · 2026-07-15: $0.22 / $0.55
input / output, per 1M tokens
Context262144
Speed/LatencyNo reliable public source
Snapshot time2026-07-15T20:11:41.551Z
SourceView source
Public activity counters

Cumulative Hugging Face model-card downloads / likes / spaces

PlatformEntryDownloadsLikesSpacesTimeSource
huggingfacegoogle/gemma-4-31B-it12,315,7653,2141002026-07-15T20:11:41.551ZView source
Operations

Runtime / access signals

TimeSignalOriginal textSource
07/13/2026, 12:10 PM UTCInconsistent Audio Support in Gemma-4-31B-IT: Video Audio Works, Standalone Audio FailsI observed an inconsistency in the multimodal behavior of google/gemma-4-31B-it when served through the OpenAI-compatible API. The model succe...Source
06/10/2026, 01:55 PM UTCChat template may re-inject prior-turn reasoning during multi-turn tool use → repetition loopsSummary The chattemplate shipped with this model may be vulnerable to a multi-turn reasoning re-injection issue that can cause verbatim repeti...Source
06/01/2026, 12:17 AM UTCRepetition collapse when continuing an assistant turn (`continue_final_message=true`)Affected models: google/gemma-4-26B-A4B-it and google/gemma-4-31B-it When continuing a partial assistant message with continuefinalmessage=tru...Source
Evidence gaps

Public display boundary

Release date not disclosedNo reliable benchmark sourceNo linkable real reviewSpeed/latency not disclosed