Skip to content

Google · Model profile

Gemma 3 12B

Intelligence, Performance & Price Analysis

Source: Canonical slug: gemma-3-12b · Canonical model registry at build time

Routing verdict

Budget-friendly / task-specific model

Best for high-volume, simple, or domain-specific tasks where cost or speed matters more than deep reasoning.

Gemma 3 12B summary metrics

Intelligence
6 scoreGemma 3 12B Instruct scores 6 on the Artificial Analysis Intelligence Index, placing it below average among other open weight non-reasoning models of similar size (median: 6).
Input Price
Not published
Output Price
Not published
Verbosity
21M Output tokens from Intelligence Index 4 out of 4 units for Verbosity . Compa

Assessment

Routing verdict and task fit

Gemma 3 12B Instruct scores 6 on the Artificial Analysis Intelligence Index, placing it below average among other open weight non-reasoning models of similar size (median: 6).
Task fit assessment
WorkloadRatingNotes
Complex reasoning & agentic workflowsSuboptimalIntelligence score 6 is better suited for straightforward tasks than multi-step reasoning.

Evidence

Evidence charts

13 published comparison charts for this profile. Gemma 3 12B is marked with a diamond and bold label; each chart has a table view and lists its source and retrieval date.

AA-Omniscience Index

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Gemma 3 12B is not marked in this published set of 20 models.
  1. Claude Fable 5 (with fallback)40.1
  2. Gemini 3.5 Flash22.7
  3. GPT-5.5 (xhigh)20.1
  4. Grok 4.3 (high)18.3
  5. Kimi K2.66.4
  6. Muse Spark4.1
  7. GLM-5.2 (max)4
  8. MiniMax-M31.4
  9. Nemotron 3 Ultra-0.833
  10. DeepSeek V4 Pro (max)-10
  11. Llama 3.1 8B-30.3
  12. gpt-oss-120b (high)-50
  13. Mistral Small 3.1-50.3
  14. Mistral Small 3.2-50.4
  15. Qwen3.6 27B-52.8
  16. Devstral Small 2-56.7
  17. Gemma 3 27B-65.9
  18. Ministral 3 14B-66.8
  19. Granite 4.1 30B-67.8
  20. Ministral 3 8B-67.8

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Artificial Analysis Intelligence Index by Open Weights / Proprietary

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Gemma 3 12B is not marked in this published set of 20 models.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. Qwen3.6 27B30.5
  12. gpt-oss-120b (high)23.8
  13. Devstral Small 217.4
  14. Mistral Small 3.114.7
  15. Ministral 3 14B11.1
  16. Mistral Small 3.210.6
  17. Ministral 3 8B9
  18. Granite 4.1 30B8.9
  19. Llama 3.1 8B7.6
  20. Gemma 3 27B7.4

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Gemma 3 12B is not marked in this published set of 20 models.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. Qwen3.6 27B30.5
  12. gpt-oss-120b (high)23.8
  13. Devstral Small 217.4
  14. Mistral Small 3.114.7
  15. Ministral 3 14B11.1
  16. Mistral Small 3.210.6
  17. Ministral 3 8B9
  18. Granite 4.1 30B8.9
  19. Llama 3.1 8B7.6
  20. Gemma 3 27B7.4

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open) · Evaluation results measured independently by Artificial Analysis

12 models in the published order. Gemma 3 12B: 50, row 3 of 12 in the published order.
  1. Nemotron 3 Ultra83.3
  2. Granite 4.1 30B61.1
  3. Gemma 3 12B (this model)50
  4. DeepSeek V4 Pro (max)50
  5. Gemma 3 27B50
  6. GLM-5.2 (max)44.4
  7. gpt-oss-120b (high)38.9
  8. Qwen3.6 27B38.9
  9. Mistral Small 3.238.9
  10. Kimi K2.633.3
  11. MiniMax-M333.3
  12. EXAONE 4.5 33B27.8

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Intelligence

Artificial Analysis Intelligence Index · Higher is better · Evaluation results measured independently by Artificial Analysis

12 models in the published order. Gemma 3 12B: 5.5, row 12 of 12 in the published order.
  1. Claude Fable 5 (with fallback)59.9
  2. GPT-5.5 (xhigh)54.8
  3. GLM-5.2 (max)51.1
  4. Gemini 3.5 Flash50.2
  5. MiniMax-M344.4
  6. DeepSeek V4 Pro (max)44.3
  7. Kimi K2.644.2
  8. Muse Spark43.1
  9. Nemotron 3 Ultra37.8
  10. Grok 4.3 (high)37.6
  11. gpt-oss-120b (high)23.8
  12. Gemma 3 12B (this model)5.5

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Speed

Output tokens per second · Higher is better · Evaluation results measured independently by Artificial Analysis

10 models in the published order. Gemma 3 12B is not marked in this published set of 10 models.
  1. gpt-oss-120b (high)316 t/s
  2. Nemotron 3 Ultra248 t/s
  3. GLM-5.2 (max)216 t/s
  4. Gemini 3.5 Flash192 t/s
  5. Grok 4.3 (high)164 t/s
  6. MiniMax-M395 t/s
  7. GPT-5.5 (xhigh)88 t/s
  8. Kimi K2.676 t/s
  9. DeepSeek V4 Pro (max)72 t/s
  10. Claude Fable 5 (with fallback)71 t/s

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Output speed context
Speed should be read together with latency and cost; fast output does not always mean better user-perceived performance.

Context Window

Context window: tokens limit · Higher is better · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Gemma 3 12B: 128k, row 18 of 20 in the published order.
  1. Gemini 3.5 Flash1M
  2. Claude Fable 5 (with fallback)1M
  3. GLM-5.2 (max)1M
  4. DeepSeek V4 Pro (max)1M
  5. Grok 4.3 (high)1M
  6. MiniMax-M31M
  7. GPT-5.5 (xhigh)922k
  8. Nemotron 3 Ultra262.1k
  9. Muse Spark262.1k
  10. Qwen3.6 27B262.1k
  11. EXAONE 4.5 33B262.1k
  12. Kimi K2.6256k
  13. Devstral Small 2256k
  14. Ministral 3 14B256k
  15. Ministral 3 8B256k
  16. gpt-oss-120b (high)131.1k
  17. Granite 4.1 30B131.1k
  18. Gemma 3 12B (this model)128k
  19. Mistral Small 3.1128k
  20. Mistral Small 3.2128k

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Context window sizing
Larger context windows enable complex multi-turn tasks but may increase latency and cost for long inputs.

Cost per Intelligence Index Task

Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better · Evaluation results measured independently by Artificial Analysis

15 models in the published order. Gemma 3 12B is not marked in this published set of 15 models.
  1. Mistral Small 3.10.00143
  2. DeepSeek V4 Pro (max)0.0063
  3. gpt-oss-120b (high)0.00454
  4. Mistral Small 3.20.00242
  5. MiniMax-M30.0139
  6. Grok 4.3 (high)0.0101
  7. Ministral 3 14B0.00235
  8. Ministral 3 8B0.00235
  9. Nemotron 3 Ultra0.0209
  10. Kimi K2.60.0216
  11. Qwen3.6 27B0.041
  12. GLM-5.2 (max)0.0268
  13. Gemini 3.5 Flash0.0884
  14. GPT-5.5 (xhigh)0.107
  15. Claude Fable 5 (with fallback)0.367

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost per Task

Weighted average cost (USD) per Intelligence Index task · Lower is better · Evaluation results measured independently by Artificial Analysis

10 models in the published order. Gemma 3 12B is not marked in this published set of 10 models.
  1. DeepSeek V4 Pro (max)$0.045
  2. gpt-oss-120b (high)$0.061
  3. MiniMax-M3$0.125
  4. Grok 4.3 (high)$0.139
  5. Nemotron 3 Ultra$0.244
  6. Kimi K2.6$0.346
  7. GLM-5.2 (max)$0.370
  8. Gemini 3.5 Flash$0.586
  9. GPT-5.5 (xhigh)$0.863
  10. Claude Fable 5 (with fallback)$2.75

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis

15 models in the published order. Gemma 3 12B is not marked in this published set of 15 models.
  1. Mistral Small 3.11.5
  2. gpt-oss-120b (high)3.5
  3. Ministral 3 8B2.8
  4. Ministral 3 14B3
  5. Mistral Small 3.22.5
  6. DeepSeek V4 Pro (max)7.2
  7. MiniMax-M314.5
  8. Grok 4.3 (high)12.5
  9. Qwen3.6 27B87.9
  10. Nemotron 3 Ultra33.5
  11. GLM-5.2 (max)32.9
  12. Kimi K2.625
  13. Gemini 3.5 Flash114
  14. GPT-5.5 (xhigh)131.8
  15. Claude Fable 5 (with fallback)507.6

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.
Cost pressure
Ethen Gateway should route away from this model when the task is simple, repetitive, or likely to generate many output tokens.

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens) · Evaluation results measured independently by Artificial Analysis

16 models; segments: Input, Output, Cache Hit. Gemma 3 12B is not marked in this published set of 16 models.
  • Input
  • Output
  • Cache Hit
  1. Llama 3.1 8B$0.170
  2. Ministral 3 8B$0.300
  3. Mistral Small 3.1$0.400
  4. Ministral 3 14B$0.400
  5. Mistral Small 3.2$0.400
  6. gpt-oss-120b (high)$0.900
  7. DeepSeek V4 Pro (max)$1.31
  8. MiniMax-M3$1.56
  9. Nemotron 3 Ultra$3.60
  10. Grok 4.3 (high)$3.95
  11. Qwen3.6 27B$4.20
  12. Kimi K2.6$5.11
  13. GLM-5.2 (max)$6.06
  14. Gemini 3.5 Flash$10.65
  15. GPT-5.5 (xhigh)$35.50
  16. Claude Fable 5 (with fallback)$61.00

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Chart interpretation
Each bar is one model measured the same way, so this model can be read against the others. The chart lists its source and retrieval date.

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference · Evaluation results measured independently by Artificial Analysis

17 models; segments: Passive Params, Active Params. Gemma 3 12B: 12.2B, row 15 of 17 in the published order.
  • Passive Params
  • Active Params
  1. DeepSeek V4 Pro (max)1,600B
  2. Kimi K2.61,000B
  3. GLM-5.2 (max)753B
  4. Nemotron 3 Ultra550B
  5. MiniMax-M3428B
  6. gpt-oss-120b (high)117B
  7. EXAONE 4.5 33B34.4B
  8. Granite 4.1 30B30B
  9. Qwen3.6 27B27.8B
  10. Gemma 3 27B27.4B
  11. Devstral Small 224B
  12. Mistral Small 3.124B
  13. Mistral Small 3.224B
  14. Ministral 3 14B14B
  15. Gemma 3 12B (this model)12.2B
  16. Ministral 3 8B8B
  17. Llama 3.1 8B8B

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Architecture scale
Larger models generally perform better but cost more to serve. Active parameter count matters for inference efficiency.

Output Tokens per Intelligence Index Task

Weighted average number of output tokens used to run one task in the Artificial Analysis Intelligence Index · Evaluation results measured independently by Artificial Analysis

20 models in the published order. Gemma 3 12B: 16,019.1, row 11 of 20 in the published order.
  1. Granite 4.1 30B3,709.5
  2. Mistral Small 3.14,757.2
  3. Gemma 3 27B5,818.7
  4. Devstral Small 26,163.9
  5. Mistral Small 3.28,068
  6. Qwen3.6 27B11,390
  7. Ministral 3 14B11,745.9
  8. Muse Spark2,857.2
  9. Grok 4.3 (high)4,032
  10. Ministral 3 8B15,686.2
  11. Gemma 3 12B (this model)16,019.1
  12. GPT-5.5 (xhigh)3,572.5
  13. Nemotron 3 Ultra7,796.7
  14. MiniMax-M311,623.4
  15. Gemini 3.5 Flash9,819.4
  16. Claude Fable 5 (with fallback)7,695.8
  17. gpt-oss-120b (high)7,571.4
  18. DeepSeek V4 Pro (max)7,243.2
  19. Kimi K2.65,397.6
  20. GLM-5.2 (max)6,089.1

Published benchmark dataset · Retrieved

Method and source
Methodology
Independent test run by Artificial Analysis on dedicated hardware.
Benchmark interpretation
Higher values indicate stronger overall capability. Bars show how models compare on composite intelligence metrics.

Specifications

Architecture and limits

Gemma 3 12B technical specifications
SpecificationValueSource
Model typeOpen weightsfaq
ReasoningNofaq
Input modalitiesGemma 3 12B Instruct supports text and image input.faq
Output modalitiesGemma 3 12B Instruct supports text output.faq
Context window130k tokensfaq
Open weights / sourceYes, Gemma 3 12B Instruct is open weights. The model weights are publicly available and can be downloaded for self-hosting.faq
ParametersGemma 3 12B Instruct has 12.2 billion parameters.faq
LicenseGemma 3 12B Instruct is released under the Gemma license license. This license allows commercial use.faq
API availabilityYes, Gemma 3 12B Instruct is available via API through 5 providers.faq

Method

Methodology & Provenance

This page is rendered from the normalized profile and page JSON for Gemma 3 12B.

Benchmark values are preserved as normalized; only layout, disclosure ordering, and typography are adjusted for readability.

Questions

Model FAQ and technical disclosures

Gemma 3 12B Instruct was released on March 12, 2025.

Gemma 3 12B Instruct was created by Google.

Gemma 3 12B Instruct scores 6 on the Artificial Analysis Intelligence Index, placing it below average among other open weight non-reasoning models of similar size (median: 6).

When evaluated on the Intelligence Index, Gemma 3 12B Instruct generated 21M output tokens, which is at the higher end compared to other open weight non-reasoning models of similar size (median: 9.5M).

No, Gemma 3 12B Instruct is not a reasoning model. It provides direct responses without extended chain-of-thought reasoning.

Gemma 3 12B Instruct supports text and image input.

Gemma 3 12B Instruct supports text output.

Yes, Gemma 3 12B Instruct supports image input and can analyze, describe, and answer questions about images.

Yes, Gemma 3 12B Instruct is multimodal. It can process text and image input and generate text output.

Gemma 3 12B Instruct has a context window of 130k tokens. This determines how much text and conversation history the model can process in a single request.

Yes, Gemma 3 12B Instruct is open weights. The model weights are publicly available and can be downloaded for self-hosting.

Gemma 3 12B Instruct has 12.2 billion parameters.

Gemma 3 12B Instruct is released under the Gemma license license. This license allows commercial use.

Gemma 3 12B Instruct achieves a score of 6 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.

Yes, Gemma 3 12B Instruct is available via API through 5 providers.

Gemma 3 12B Instruct is available through 5 API providers.

Models

  • 01GoogleOther Google models in the catalog.
  • 02LeaderboardsCertified metric ranks, when published.