Noometry

Google, open weights

Gemma 3 12B

Gemma 3 12B by Google ranks 262nd of 354 ranked models on the Noometry Index as of October 2026, with a score of 32.1. Its strongest category is agentic & tool use, where it ranks 108th. API pricing starts at $0.05 per million input tokens and $0.15 per million output tokens, with a 131K-token context window.

Last verified

Noometry rank
#262 of 354
Index score
32.1
Evidence
Confirmed 24 results
Provider
Google
Released
March 12, 2025
Weights
Open weights
Reasoning
No
Context window
131K
Max output
8K
Input price
$0.05 / M
Output price
$0.15 / M
Blended price
$0.075 / M
Output speed
Not measured
Value
#11 of 219
Knowledge cutoff
August 2024
Input
text, image
Hugging Face
google/gemma-3-12b-it

Category scores

Each category score combines every public result we have in that category.

  1. Coding 31.7
  2. Agentic & Tool Use 25.5
  3. Reasoning 15.7
  4. Math 22.3
  5. Knowledge 26.5
  6. Multilingual 45.7
  7. Instruction Following 68.6
  8. Long Context 40.0
  9. Writing & Preference 47.5
CategoryScoreRankResults
Coding31.7#2802
Agentic & Tool Use25.5#1081
Reasoning15.7#3135
Math22.3#2792
Knowledge26.5#2573
Multilingual45.7#1651
Instruction Following68.6#1861
Long Context40.0#1621
Writing & Preference47.5#2094

Strengths and weaknesses

Categories where Gemma 3 12B places highest and lowest among the models ranked in each, with its score against that category's median.

Strongest categories

CategoryScorevs medianRank
Long Context40.0−0.9#162 of 296, top 55%
Multilingual45.7−1.7#165 of 297, top 56%
Instruction Following68.6−2.7#186 of 305, top 61%

Weakest categories

CategoryScorevs medianRank
Reasoning15.7−7.9#313 of 350, top 90%
Math22.3−14.3#279 of 327, top 86%
Coding31.7−7.0#280 of 340, top 83%

Closest competitors

The models ranked just above and below Gemma 3 12B. When scores are this close, price and speed are often the better way to choose.

ModelRankScoreBlended $/MSpeed
Granite 3.1 8b Instruct#25832.4——Compare
Pixtral Large#25932.2$3—Compare
Falcon-180B#26032.2——Compare
Gemini 1.5 Pro (May 2024)#26132.1——Compare
Mistral Large#26331.9$3—Compare
Qwen3-4B#26431.9——Compare
Amazon Nova Lite#26531.9$0.10—Compare
Mistral Medium 3.1#26631.9$0.80—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

BenchmarkScorePositionSettingSourceDate
SciCode17.4%#118 of 121, top 98%Epoch AI
LMArena Coding1281#210 of 294, top 72%LMArena2026-10-08

Agentic & Tool Use

BenchmarkScorePositionSettingSourceDate
Berkeley Function Calling Leaderboard30.4%#33 of 49, top 68%promptBerkeley Function Calling Leaderboard

Reasoning

BenchmarkScorePositionSettingSourceDate
CritPt0%#107 of 134, top 80%Epoch AI
Chess Puzzles0%#110 of 129, top 86%Epoch AI2026-08-28
LMArena Hard Prompts1309#183 of 297, top 62%LMArena2026-10-08
DTBench48.8%#139 of 151, top 93%Epoch AI
LMCA4.5%#124 of 125, top 100%Epoch AI
Epoch Capabilities Index123.5#160 of 213, top 76%Epoch AI2025-03-12

Math

BenchmarkScorePositionSettingSourceDate
OTIS Mock AIME 2024-202516.7%#128 of 173, top 74%Epoch AI2026-08-28
LMArena Math1307#178 of 285, top 63%LMArena2026-10-08

Knowledge

BenchmarkScorePositionSettingSourceDate
GPQA Diamond39.5%#151 of 186, top 82%Epoch AI2026-08-28
Vectara Hallucination Rate (lower is better)4.4%#5 of 96, top 6%Vectara Hallucination Leaderboard
LMArena Expert1248#198 of 273, top 73%LMArena2026-10-08

Multimodal

BenchmarkScorePositionSettingSourceDate
MindCube46.7%Best of 2Epoch AI

Multilingual

BenchmarkScorePositionSettingSourceDate
LMArena Non-English1318#165 of 297, top 56%LMArena2026-10-08
LMArena German1370#116 of 231, top 51%LMArena2026-10-08
LMArena Russian1335#155 of 283, top 55%LMArena2026-10-08

Instruction Following

BenchmarkScorePositionSettingSourceDate
LMArena Instruction Following1299#179 of 298, top 61%LMArena2026-10-08

Long Context

BenchmarkScorePositionSettingSourceDate
LMArena Longer Query1317#174 of 291, top 60%LMArena2026-10-08

Writing & Preference

BenchmarkScorePositionSettingSourceDate
LMArena Text1334#170 of 297, top 58%LMArena2026-10-08
LMArena Creative Writing1331#145 of 295, top 50%LMArena2026-10-08
EQ-Bench Creative Writing1126#92 of 115, top 80%EQ-Bench
LMArena Multi-Turn1334#169 of 295, top 58%LMArena2026-10-08

API pricing by provider

RouteInput $/MOutput $/MCached input $/MChecked
bedrock$0.09$0.29—2026-10-10
deepinfra$0.05$0.15—2026-10-10
openrouter$0.05$0.15—2026-10-10

All Google API prices → Estimate your cost →

Compare Gemma 3 12B

Other Google models

Frequently asked questions

How good is Gemma 3 12B?

Gemma 3 12B by Google ranks 262nd of 354 ranked models on the Noometry Index as of October 2026, with a score of 32.1. Its strongest category is agentic & tool use, where it ranks 108th. API pricing starts at $0.05 per million input tokens and $0.15 per million output tokens, with a 131K-token context window.

How much does Gemma 3 12B cost?

Gemma 3 12B costs $0.05 per million input tokens and $0.15 per million output tokens on deepinfra.

What is Gemma 3 12B's context window?

Gemma 3 12B accepts up to 131K tokens of input and can write up to 8K tokens in one response.

Is Gemma 3 12B open source?

Yes. Gemma 3 12B's weights are downloadable from Hugging Face (google/gemma-3-12b-it); check the license for commercial terms.

What are Gemma 3 12B's strengths and weaknesses?

Relative to other ranked models, Gemma 3 12B places best in long context, multilingual, instruction following and lowest in reasoning, math, coding.

What is Gemma 3 12B best at?

Its best category is agentic & tool use, where it ranks 108th on Noometry.