Venice Atlas
← Back to model catalog

VENICE ATLAS · MODEL DOSSIER · GOOGLE GEMMA 3 27B INSTRUCT

Model overview

Google Gemma 3 27B Instruct

Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities, including structured outputs and function calling. Gemma 3 27B is Google's latest open source model, successor to Gemma 2.

google-gemma-3-27b-it · venice.ai
Text · Venice catalog$0.12 per 1M tokensPrice on Venice AI
ModalityTextOn Venice AI
Primary price$0.12 per 1M tokensper 1M tokens
Context window198KMax tokens
Benchmarks27Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

Chat#1021.67 ratingFactuality#68.87 ratingGrounding#1014.06 ratingInstruction Following#1121.67 ratingStructured Output#1421.67 rating
Chat1 result
IFEvalifeval · chat
90.4%
ReportedOct 2026
Factuality1 result
SimpleQAsimpleqa · factuality
10.0%
ReportedOct 2026
General intelligence1 result
Global-MMLU-Liteglobal-mmlu-lite · general
75.1%
ReportedOct 2026
Language3 results
ECLeKTiceclektic · language
16.7%
ReportedOct 2026
MMLU-Prommlu-pro · language
67.5%
ReportedOct 2026
WMT24++wmt24++ · language
53.4%
ReportedOct 2026
Long Context1 result
MRCR v2 (8-needle)mrcr-v2-(8-needle) · long_context
13.5%
ReportedOct 2026
Math4 results
GSM8kgsm8k · math
95.9%
ReportedOct 2026
HiddenMathhiddenmath · math
60.3%
ReportedOct 2026
MATHmath · math
89.0%
ReportedOct 2026
MathVista-Minimathvista-mini · math
67.6%
ReportedOct 2026
Reasoning10 results
BIG-Bench Extra Hardbig-bench-extra-hard · reasoning
19.3%
ReportedOct 2026
BIG-Bench Hardbig-bench-hard · reasoning
87.6%
ReportedOct 2026
Bird-SQL (dev)bird-sql-(dev) · reasoning
54.4%
ReportedOct 2026
ChartQAchartqa · reasoning
78.0%
ReportedOct 2026
FACTS Groundingfacts-grounding · reasoning
74.9%
ReportedOct 2026
GPQAgpqa · reasoning
42.4%
ReportedOct 2026
HumanEvalhumaneval · reasoning
87.8%
ReportedOct 2026
LiveCodeBenchlivecodebench · reasoning
29.7%
ReportedOct 2026
MBPPmbpp · reasoning
74.4%
ReportedOct 2026
Natural2Codenatural2code · reasoning
84.5%
ReportedOct 2026
Vision6 results
AI2Dai2d · vision
84.5%
ReportedOct 2026
DocVQAdocvqa · vision
86.6%
ReportedOct 2026
InfoVQAinfovqa · vision
70.6%
ReportedOct 2026
MMMU (val)mmmu-(val) · vision
64.9%
ReportedOct 2026
TextVQAtextvqa · vision
65.1%
ReportedOct 2026
VQAv2 (val)vqav2-(val) · vision
71.0%
ReportedOct 2026

Learn what each benchmark measures ↗

Related models

More from google

gemini-3-8-flashGemini 3.8 Flash$0.94 input per 1M tokens ↗gemini-4-argonGemini 4 Argon$2.50 input per 1M tokens ↗gemini-3-1-pro-previewGemini 3.1 Pro Preview$2.50 input per 1M tokens ↗