Skip to main content

Model profile

Gemma 4 12B

GoogleCurrentReleased Jun 3, 2026
Data verified
Overall Score
47.29Public #137 of 200
Arena Elo
Not listed
Eligible category ranks
0of 8
Price (1M tokens)
Not listedAPI pricing
Speed
Not listed
Context
256K

Evidence coverage

23 of 323 tracked benchmarks are published. 11 are verified and 12 provisional. 7 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
23 / 323
Verified
11
Provisional
12
Categories with evidence
7 / 8

Evidence by category

  • Agentic1 benchmark
    Reported
  • Coding1 benchmark
    Reported
  • Reasoning4 benchmarks
    Mixed evidence
  • Knowledge11 benchmarks
    Mixed evidence
  • Math1 benchmark
    Verified
  • Multilingual0 benchmarks
    Not measured
  • Multimodal4 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
Open WeightSelf-hostReasoning
Confidence:
Medium
12b

Gemma 4 12B ranks #137 out of 200 models on the public leaderboard with an overall score of 47.29/100. It does not yet have enough sourced coverage for BenchLM's verified leaderboard. While not a frontier model, it offers specific advantages depending on the use case.

Gemma 4 12B is a open weight model with a 256K token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

Gemma 4 12B sits inside the Gemma 4 family alongside Gemma 4 31B, Gemma 4 26B A4B, Gemma 4 E2B, Gemma 4 E4B. This profile currently has 23 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 47.247.74

  1. Claude 3.5 Sonnet
    Anthropic
    #13247.74
    Claude 3.5 Sonnet is #132 with a score of 47.74.
    Compare
  2. GLM-4.5-Air
    Z.AI
    #13347.7
    GLM-4.5-Air is #133 with a score of 47.7.
    Compare
  3. Qwen3.5 Flash
    Alibaba
    #13447.65
    Qwen3.5 Flash is #134 with a score of 47.65.
    Compare
  4. Command A+
    Cohere
    #13547.51
    Command A+ is #135 with a score of 47.51.
    Compare
  5. o3-mini
    OpenAI
    #13647.41
    o3-mini is #136 with a score of 47.41.
    Compare
  6. Gemma 4 12BCurrent model
    Google
    #13747.29
    Gemma 4 12B is #137 with a score of 47.29.
  7. Qwen3.5 Plus
    Alibaba
    #13847.2
    Qwen3.5 Plus is #138 with a score of 47.2.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

No eligible category percentile is available from the published evidence yet.

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticWeight 22%1 benchmarkReportedScore pending
CodingRank Not rankedWeight 20%1 benchmarkReported72.0
ReasoningRank Not rankedWeight 17%4 benchmarksMixed sources24.0
KnowledgeRank Not rankedWeight 12%11 benchmarksMixed sources70.5
MathRank Not rankedWeight 5%1 benchmarkVerified57.7
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank Not rankedWeight 12%4 benchmarksMixed sources27.7
Inst. FollowingWeight 5%1 benchmarkReportedScore pending

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic1 benchmark
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

36.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Coding1 benchmark
AA-SciCodeReported

Artificial Analysis SciCode

38.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Reasoning4 benchmarks
MRCRv2Provider exact
43.4%Weighted 31%
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Google reports MRCR v2 8-needle 128k average at 43.4.
BBHProvider exact

BIG-Bench Hard

53%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
AA-LCRReported

Artificial Analysis Long Context Reasoning

55.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

0.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Knowledge11 benchmarks
MMLU-ProProvider exact

Massive Multitask Language Understanding Professional

77.2%Weighted 30%
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
GPQAProvider exact

Graduate-Level Google-Proof Q&A

78.8%Weighted 7%
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Google reports GPQA Diamond at 78.8. BenchLM stores that exact value on the weighted GPQA lane and the display GPQA-Diamond lane.
GPQA-DProvider exact

GPQA Diamond

78.8%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Google reports GPQA Diamond at 78.8.
HLE w/o toolsProvider exact

Humanity's Last Exam without tools

5.2%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Google reports HLE no tools at 5.2; no with-search score is published for Gemma 4 12B in this table.
MMMLUProvider exact
83.4%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
Artificial Analysis Intelligence IndexReported
22.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

75.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

14.8%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

-51.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

16.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

80.8%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math1 benchmark
AIME26Provider exact

AIME 2026

77.5%Weighted 25%
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Google reports AIME 2026 no tools at 77.5.
Multimodal4 benchmarks
MMMU-ProProvider exact

Massive Multi-discipline Multimodal Understanding Pro

69.1%Weighted 45%
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
MathVisionProvider exact
79.7%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
MedXpertQA (MM)Provider exact

MedXpertQA Multimodal

48.7%Display only
Source: Google Gemma 4 12B Hugging Face model cardProvenance: Provider exact
AA-MMMU-ProReported

Artificial Analysis MMMU-Pro

69.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

73.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

Frequently Asked Questions

How does Gemma 4 12B perform overall in AI benchmarks?

Gemma 4 12B has 23 published benchmark scores on BenchLM, but it does not yet have enough non-generated coverage to receive a global overall rank.

Is Gemma 4 12B good for knowledge and understanding?

Gemma 4 12B has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for coding and programming?

Gemma 4 12B has visible benchmark coverage in coding and programming, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for mathematics?

Gemma 4 12B has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for reasoning and logic?

Gemma 4 12B has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for agentic tool use and computer tasks?

Gemma 4 12B has visible benchmark coverage in agentic tool use and computer tasks, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for multimodal and grounded tasks?

Gemma 4 12B has visible benchmark coverage in multimodal and grounded tasks, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B good for instruction following?

Gemma 4 12B has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Is Gemma 4 12B open source?

Yes, Gemma 4 12B is an open weight model created by Google, meaning it can be downloaded and run locally or fine-tuned for specific use cases.

Which sibling models are related to Gemma 4 12B?

Gemma 4 12B belongs to the Gemma 4 family. Related variants on BenchLM include Gemma 4 31B, Gemma 4 26B A4B, Gemma 4 E2B, Gemma 4 E4B.

Does Gemma 4 12B have full benchmark coverage on BenchLM?

Not yet. Gemma 4 12B currently has 23 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Gemma 4 12B?

Gemma 4 12B has a published context window of 256K, which determines how much text it can process in a single interaction.

Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.