Google, open weights
Gemma 3 27B
Gemma 3 27B by Google ranks 284th of 354 ranked models on the Noometry Index as of October 2026, with a score of 30.8. Its strongest category is multimodal, where it ranks 100th. API pricing starts at $0.08 per million input tokens and $0.16 per million output tokens, with a 131K-token context window.
Last verified
Specifications
- Noometry rank
- #284 of 354
- Index score
- 30.8
- Evidence
- Confirmed 43 results
- Provider
Google- Released
- March 11, 2025
- Weights
- Open weights
- Reasoning
- No
- Context window
- 131K
- Max output
- 8K
- Input price
- $0.08 / M
- Output price
- $0.16 / M
- Blended price
- $0.10 / M
- Output speed
- 62 tokens/s Kagi
- Value
- #17 of 219
- Knowledge cutoff
- August 2024
- Input
- text, image
- Hugging Face
- google/gemma-3-27b-it
Category scores
Each category score combines every public result we have in that category.
- Coding 22.5
- Agentic & Tool Use 25.1
- Reasoning 16.7
- Math 25.9
- Knowledge 25.5
- Multimodal 32.6
- Multilingual 46.9
- Instruction Following 70.6
- Long Context 27.6
- Writing & Preference 52.5
| Category | Score | Rank | Results |
|---|---|---|---|
| Coding | 22.5 | #334 | 4 |
| Agentic & Tool Use | 25.1 | #110 | 1 |
| Reasoning | 16.7 | #301 | 8 |
| Math | 25.9 | #265 | 4 |
| Knowledge | 25.5 | #261 | 4 |
| Multimodal | 32.6 | #100 | 2 |
| Multilingual | 46.9 | #155 | 1 |
| Instruction Following | 70.6 | #160 | 2 |
| Long Context | 27.6 | #293 | 2 |
| Writing & Preference | 52.5 | #168 | 6 |
Strengths and weaknesses
Categories where Gemma 3 27B places highest and lowest among the models ranked in each, with its score against that category's median.
Strongest categories
| Category | Score | vs median | Rank |
|---|---|---|---|
| Multilingual | 46.9 | −0.6 | #155 of 297, top 53% |
| Instruction Following | 70.6 | −0.6 | #160 of 305, top 53% |
| Writing & Preference | 52.5 | −1.3 | #168 of 312, top 54% |
Weakest categories
| Category | Score | vs median | Rank |
|---|---|---|---|
| Long Context | 27.6 | −13.3 | #293 of 296, top 99% |
| Coding | 22.5 | −16.2 | #334 of 340, top 99% |
| Reasoning | 16.7 | −6.9 | #301 of 350, top 86% |
Closest competitors
The models ranked just above and below Gemma 3 27B. When scores are this close, price and speed are often the better way to choose.
| Model | Rank | Score | Blended $/M | Speed | |
|---|---|---|---|---|---|
| Mistral Small 3.2 | #280 | 31.2 | $0.13 | 68 | Compare |
| Amazon Nova Pro | #281 | 31.0 | $1.40 | — | Compare |
| Llama 4 Maverick | #282 | 30.9 | $0.30 | 456 | Compare |
| Phi-4 Mini | #283 | 30.9 | $0.13 | — | Compare |
| Qwen1.5-72B | #285 | 30.8 | — | — | Compare |
| Granite 3.0 2b Instruct | #286 | 30.8 | — | — | Compare |
| Codellama 34b Instruct | #287 | 30.8 | — | — | Compare |
| Llama 3.1-405B | #288 | 30.7 | — | 78 | Compare |
Sponsored placements are available on pages like this one. Advertise on Noometry
Benchmark results
Every published result we track, with its source. Bold rows are the ones used for ranking; where several exist we prefer independent runs over self-reported numbers.
Coding
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| Aider Polyglot | 4.9% | #43 of 44, top 98% | Epoch AI | ||
| SciCode | 21.2% | #117 of 121, top 97% | Epoch AI | ||
| LiveBench Coding | 39.9% | #23 of 39, top 59% | Epoch AI | ||
| LMArena Coding | 1322 | #183 of 294, top 63% | LMArena | 2026-10-08 |
Agentic & Tool Use
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| Berkeley Function Calling Leaderboard | 29.5% | #34 of 49, top 70% | prompt | Berkeley Function Calling Leaderboard |
Reasoning
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| Kagi LLM Benchmark | 35.1% | Kagi LLM Benchmark | |||
| Kagi LLM Benchmark | 40.4% | #78 of 99, top 79% | Kagi LLM Benchmark | ||
| CritPt | 0% | #108 of 134, top 81% | Epoch AI | ||
| Chess Puzzles | 0% | #112 of 129, top 87% | Epoch AI | 2026-08-27 | |
| LiveBench Reasoning | 43.8% | #24 of 39, top 62% | Epoch AI | ||
| LMArena Hard Prompts | 1340 | #164 of 297, top 56% | LMArena | 2026-10-08 | |
| DTBench | 52.5% | #129 of 151, top 86% | Epoch AI | ||
| LiveBench Data Analysis | 51.5% | #21 of 39, top 54% | Epoch AI | ||
| LMCA | 12.3% | #108 of 125, top 87% | Epoch AI | ||
| Epoch Capabilities Index | 130.04 | #138 of 213, top 65% | Epoch AI | 2025-03-12 | |
| LiveBench | 50% | #20 of 39, top 52% | Epoch AI |
Math
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 22.5% | #124 of 173, top 72% | Epoch AI | 2026-08-27 | |
| LiveBench Math | 55.4% | #17 of 39, top 44% | Epoch AI | ||
| LMArena Math | 1312 | #175 of 285, top 62% | LMArena | 2026-10-08 | |
| MATH Level 5 | 74% | #28 of 79, top 36% | Epoch AI | 2025-03-13 |
Knowledge
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| GPQA Diamond | 47.7% | #133 of 186, top 72% | Epoch AI | 2026-08-27 | |
| Confabulations (lower is better) | 40.3% | #51 of 51, top 100% | Lech Mazur benchmarks | ||
| Vectara Hallucination Rate (lower is better) | 7.4% | #31 of 96, top 33% | Vectara Hallucination Leaderboard | ||
| LMArena Expert | 1304 | #177 of 273, top 65% | LMArena | 2026-10-08 |
Multimodal
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Vision | 1164 | #93 of 122, top 77% | LMArena | 2026-10-09 | |
| GeoBench | 52% | #18 of 25, top 72% | Epoch AI |
Multilingual
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Non-English | 1334 | #155 of 297, top 53% | LMArena | 2026-10-08 | |
| LMArena Chinese | 1346 | #169 of 285, top 60% | LMArena | 2026-10-08 | |
| LMArena French | 1368 | #137 of 223, top 62% | LMArena | 2026-10-08 | |
| LMArena German | 1362 | #122 of 231, top 53% | LMArena | 2026-10-08 | |
| LMArena Japanese | 1287 | #127 of 211, top 61% | LMArena | 2026-10-08 | |
| LMArena Korean | 1308 | #127 of 213, top 60% | LMArena | 2026-10-08 | |
| LMArena Russian | 1349 | #148 of 283, top 53% | LMArena | 2026-10-08 | |
| LMArena Spanish | 1349 | #145 of 226, top 65% | LMArena | 2026-10-08 |
Instruction Following
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LiveBench Instruction Following | 74.9% | #15 of 39, top 39% | Epoch AI | ||
| LMArena Instruction Following | 1321 | #161 of 298, top 55% | LMArena | 2026-10-08 |
Long Context
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| Fiction.LiveBench | 33.3% | #45 of 47, top 96% | Epoch AI | ||
| LMArena Longer Query | 1333 | #161 of 291, top 56% | LMArena | 2026-10-08 |
Writing & Preference
| Benchmark | Score | Position | Setting | Source | Date |
|---|---|---|---|---|---|
| LMArena Text | 1358 | #152 of 297, top 52% | LMArena | 2026-10-08 | |
| LMArena Creative Writing | 1346 | #133 of 295, top 46% | LMArena | 2026-10-08 | |
| Short-Story Creative Writing | 79.9% | #16 of 39, top 42% | Epoch AI | ||
| EQ-Bench Creative Writing | 1266 | #81 of 115, top 71% | EQ-Bench | ||
| LMArena Multi-Turn | 1345 | #156 of 295, top 53% | LMArena | 2026-10-08 | |
| LiveBench Language | 34.6% | #23 of 39, top 59% | Epoch AI |
API pricing by provider
| Route | Input $/M | Output $/M | Cached input $/M | Checked |
|---|---|---|---|---|
| bedrock | $0.23 | $0.38 | — | 2026-10-10 |
| deepinfra | $0.08 | $0.16 | — | 2026-10-10 |
| openrouter | $0.08 | $0.45 | $0.04 | 2026-10-10 |
Compare Gemma 3 27B
- Gemma 3 27B vs Gemma 2 27B
- Gemma 3 27B vs Phi-4 Mini
- Gemma 3 27B vs Qwen1.5-72B
- Gemma 3 27B vs Llama 4 Maverick
- Gemma 3 27B vs Granite 3.0 2b Instruct
- Gemma 3 27B vs Amazon Nova Pro
- Gemma 3 27B vs Codellama 34b Instruct
- Gemma 3 27B vs GPT-6 Astra
- Gemma 3 27B vs Claude Fable 5.1
- Gemma 3 27B vs Kimi K3
- Gemma 3 27B vs Grok 4.6
- Gemma 3 27B vs Qwen3.8 Max
- Gemma 3 27B vs GLM-5.3
- Gemma 3 27B vs Muse Spark 1.3
Other Google models
Frequently asked questions
How good is Gemma 3 27B?
Gemma 3 27B by Google ranks 284th of 354 ranked models on the Noometry Index as of October 2026, with a score of 30.8. Its strongest category is multimodal, where it ranks 100th. API pricing starts at $0.08 per million input tokens and $0.16 per million output tokens, with a 131K-token context window.
How much does Gemma 3 27B cost?
Gemma 3 27B costs $0.08 per million input tokens and $0.16 per million output tokens on deepinfra.
What is Gemma 3 27B's context window?
Gemma 3 27B accepts up to 131K tokens of input and can write up to 8K tokens in one response.
Is Gemma 3 27B open source?
Yes. Gemma 3 27B's weights are downloadable from Hugging Face (google/gemma-3-27b-it); check the license for commercial terms.
How fast is Gemma 3 27B?
Gemma 3 27B generated about 62 output tokens per second in the Kagi LLM Benchmark's timed runs. Speed varies by provider, load and reasoning effort.
What are Gemma 3 27B's strengths and weaknesses?
Relative to other ranked models, Gemma 3 27B places best in multilingual, instruction following, writing & preference and lowest in long context, coding, reasoning.
What is Gemma 3 27B best at?
Its best category is multimodal, where it ranks 100th on Noometry.