NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Gemma 2 27B
Google DeepMind logo

Gemma 2 27B

by Google DeepMind · Released Jul 2024

Open Source
50.1
avg score
Rank #144
Compare
Better than 54% of all models
Context
8K tokens (~4 books)
Input $/1M
$0.65
Output $/1M
$0.65
Type
text
License
Open Source
Benchmarks
15 tested
Data as of
About

Gemma 2 27B by Google is an open model built from the same research and technology used to create the Gemini models. Gemma models are well-suited for a variety of...

Tested on 15 benchmarks · BenchGecko score 50.1. Top scores: Chatbot Arena Elo — Overall (1289.4%), GSM8K (84.9%), IFEval (79.8%).

Looking for similar performance at lower cost?
Llama 3.1 70B Instruct scores 49.7 (99% as good) at $0.40/1M input · 38% cheaper
Capabilities
reasoning
9.1
#185 globally
math
34.5
#176 globally
knowledge
41.1
#198 globally
general
23.7
#149 globally
language
79.8
#58 globally
Benchmark Scores
Compare All
Tested on 15 benchmarks · Ranked across 6 categories
Score Distribution (all 312 models)
0255075100
▲ You are here
reasoningCompare reasoning →
MUSR

HuggingFace MuSR (Multi-Step Reasoning). Tests multi-hop reasoning requiring chaining multiple facts together.

9.1·
mathCompare math →
GSM8K

Grade school math word problems. 8,500 problems testing multi-step arithmetic reasoning. A foundational math benchmark.

84.9·
MATH level 5

Competition-level math from AMC, AIME, and olympiad problems. Level 5 is the hardest tier, requiring creative problem-solving.

27.9·
MATH Level 5

HuggingFace evaluation of MATH Level 5 problems. Competition math requiring advanced reasoning and proof construction.

23.9·
knowledgeCompare knowledge →
MMLU

Massive Multitask Language Understanding. 57 subjects from STEM, humanities, and social sciences. The most widely-cited knowledge benchmark.

67.6·
PIQA

Physical Intuition QA. Tests understanding of everyday physical interactions and commonsense physics.

67.4·
MMLU-PRO

HuggingFace MMLU-Pro. Harder version of MMLU with 10 answer choices instead of 4 and more challenging questions.

38.4·
Quick compare:Mistral AI logovs Mistral 7B V0.1TII logovs Falcon-180BAnthropic logovs Claude Opus 4
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Similar Models
Mistral AI logo
Mistral 7B V0.1
Mistral AI
50.0n/a
TII logo
Falcon-180B
TII
50.3n/a
Anthropic logo
Claude Opus 4
Anthropic
49.8$15.00/1M
Links
Info
Google DeepMindPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@Google DeepMind
Source Code
GitHubHugging Face
BenchGecko API
gemma-2-27b-it
Specifications
  • Typetext
  • Context8K tokens (~4 books)
  • ReleasedJul 2024
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.002
Available On
Google DeepMind logoGoogle DeepMind$0.65
Categories
reasoningmathknowledgegenerallanguage
Learn More
transformeropen weightstokens
Share & Export
Tweet
Related Models
Mistral AI logoMistral 7B V0.1
50.0
TII logoFalcon-180B
50.3
Anthropic logoClaude Opus 4
49.8
Meta logoLlama 3.1 70B Instruct
49.7
Gemma 2 27B is an open-source text AI model by Google DeepMind, released in July 2024. It has an average benchmark score of 50.1. Context window: 8K tokens.

Related Models

Mistral 7B V0.1 · Mistral AIFalcon-180B · TIIClaude Opus 4 · AnthropicLlama 3.1 70B Instruct · MetaGemini 2.0 Pro · Google DeepMind

Benchmarks

Chatbot Arena Elo — OverallGSM8KIFEvalMMLUPIQA

Related Pages

Google DeepMind · ProviderGoogle DeepMind · EconomyAll ModelsCompare ModelsPricingDevelopers · API

Key facts · as of 2026-10-05

  • Gemma 2 27B by Google DeepMind. BenchGecko score 50.1, rank 144 of 312 scored models (normalized average of public benchmark scores).
  • List price $0.65 input · $0.65 output per 1M tokens (as of 2026-10-05).
  • Sold by 1 provider (as of 2026-10-05): NextBit (int4) $0.65 in / $0.65 out.

How to cite · data as of 2026-10-05

Gemma 2 27B · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/gemma-2-27b-it

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP