Gemma 3 27B
google/gemma-3-27b-it- Family · Gemma 3
- Open weights
- Text · Image → Text
- Tools · JSON · Vision
Not on OSMEXO yet
Spec & list price
Context128K
Max output118K
Input / 1M$0.08
Output / 1M$0.354.4× input
Usage
Tokens / day8.5B▲ 1.4% 7d
Usage rank#94
Requests / day2.8M
Tool calls / req0.00
Per request · 30 days
Avg input1.3K0.1× median
Avg output3960.6× median
Cache hit28%0.4× median
Tool-call failures9%4.7× median
Consumption
Peak 15.2B 5 OctLow 2.8B 17 OctRank today #94
News & trends
2 mentions this weekGemma 3 family · videos, posts, articles · 30 days
News · latest from Google
- Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v529 Oct
- Gemini 4 Carbon (High) Scores 59 on Artificial Analysis v528 Oct
- Gemini 4 Carbonblog.google · 28 Oct
- Gemini 4 Carbon: Our next frontier modelGoogle DeepMind · 28 Oct
Benchmarks
Benchmark data: Artificial AnalysisOutput speed#301/33429 tok/s2.3× slower than median
Fastest provider37 tok/sbest of 5 providers
First token#2011.32 s1.3× longer than median
Family progress
Intelligence Index
- Family releases5.1
Technical details
- Tokenizer
- Gemini
- Knowledge cutoff
- 2024-08-31
- License
- gemma
- Parameters
- 27.4B
- Moderation
- none
- Downloads · likes
- 351.3K · 2.1K
Request parameters 17
- tools
- tool_choice
- response_format
- structured_outputs
- temperature
- top_p
- top_k
- min_p
- max_tokens
- stop
+7 more
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
Interesting facts
- The most used of 3 Gemma 3 models by tokens last month.
- Weekdays average 6.3B tokens a day against 7.0B on weekends.
When to use
Best for
- Self-hosted chat with image input
- Private deployments where data stays in-house
- Fine-tuning a capable open model
- Document and screenshot Q&A
Think twice for
- Very long documents (limited context)
- Hardest reasoning and coding tasks
- Audio or video input
Research papers
- Coverage, Not Difficulty, Sets How Much Synthetic Data an Activation Probe Needs6 Oct
- Feature Elicitation for SAE Steering of Deceptive Behaviour in Small LLMs2 Oct
- AVSD-Scenes: A Dataset for Audio-Visual Description of Urban Scenes1 Oct
- TeochewBench: A Human-Reviewed Benchmark for Teochew Hanzi Translation16 Sep
- Investigating first-language bias in LLM-based automated essay scoring: A cross-prompt evaluation of an open-weight AI-model on TOEFL essays16 Jul
- The Illusion of Safety: Multi-Tier Verification of AI vs. Human C++ Code30 Jun
- Internal-State Probes Read the Situation, Not the Action: Three Negative Results for Pre-Action Misalignment Monitoring29 Jun
- Benchmarking Large Language Models on Reference Extraction and Parsing in the Social Sciences and Humanities13 Mar
Same family
Gemma 3- 4.8
- 5.1
- 4.8
- 3.8
Want Gemma 3 27B on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.