Gemma 4 12B Unified Benchmarks, Specifications & Availability

Gemma 4 12B Unified

Explore Gemma 4 12B Unified from Google: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Published specifications

Provider: Google
Access: Open weights
License: Apache-2.0
Context window: 256K
Total parameters: 11.95B
Active parameters: Not published
Released: 2026-06-03
Modalities: text, image, video, audio
Family: Gemma 4

Hugging Face · Announcement · Website

Model notes

Public ungated Apache-2.0 instruction-tuned checkpoint; PT, assistant/MTP and QAT builds are not extra curated models. Vendor architecture size is retained separately from HF tensor totals including modality modules. 12B Unified projects raw image patches/audio into one decoder and has no separate modality encoders. Audio input is supported for up to30 seconds; video uses frames (up to60 seconds at1fps), text output only. Context is the published configured limit, not a universal guarantee of whole-window accuracy. Reasoning is an on/off mode, not an API effort level. Report Table9 disables thinking; vision uses1120 or280 visual-token budgets. OmniDocBench uses native average edit distance, not accuracy. Card/report audio aggregates and the older launch τ² figures differ and remain source-specific observations. Released3June, separate from the March family launch. No exact public OpenRouter route found19October.

Compare Gemma 4 12B Unified with other models →Compare subscription plans Explore data coverage

Pricing · OpenRouter

No OpenRouter pricing is recorded for this model. Missing rates are not free.

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
MMLU-Pro	77.2%unknown effort	All 1 recorded result & sources 77.2% · raw 77.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
AIME 2026	77.5%unknown effort	All 1 recorded result & sources 77.5% · raw 77.5 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. No tools. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
LiveCodeBench v6	72%unknown effort	All 1 recorded result & sources 72% · raw 72 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Codeforces Rating	1659unknown effort	All 1 recorded result & sources 1659 · raw 1659 rating Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
SciCode (Gemma report; aggregation unspecified)	38%unknown effort	All 1 recorded result & sources 38% · raw 38 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
GPQA Diamond	78.8%unknown effort	All 1 recorded result & sources 78.8% · raw 78.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
BigBench Extra Hard (micro average)	53%unknown effort	All 1 recorded result & sources 53% · raw 53 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Humanity's Last Exam (no tools)	5.2%unknown effort	All 1 recorded result & sources 5.2% · raw 5.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
IFBench (Gemma report protocol)	74%unknown effort	All 1 recorded result & sources 74% · raw 74 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. IFBench strict/loose convention is not stated in this table. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
IFEval	97.2%unknown effort	All 1 recorded result & sources 97.2% · raw 97.2 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MMMLU	83.4%unknown effort	All 1 recorded result & sources 83.4% · raw 83.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MRCR v2 (8 needles,128K average)	43.4%unknown effort	All 1 recorded result & sources 43.4% · raw 43.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
Terminal Bench Hard (Gemma report)	18%unknown effort	All 1 recorded result & sources 18% · raw 18 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Airline)	75%unknown effort	All 1 recorded result & sources 75% · raw 75 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Retail)	77.6%unknown effort	All 1 recorded result & sources 77.6% · raw 77.6 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
τ²-bench (Telecom)	54.4%unknown effort	All 1 recorded result & sources 54.4% · raw 54.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google final instruction-tuned report Table5; thinking enabled, no effort tier inferred. Report dated19June2026, not execution date. Visually checked numeric table. https://arxiv.org/pdf/2607.02770#page=5
MMMU Pro (no tools)	69.1%unknown effort	All 1 recorded result & sources 69.1% · raw 69.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MMMU Pro (Gemma,280 visual tokens)	67.7%unknown effort	All 1 recorded result & sources 67.7% · raw 67.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
MATH-Vision	79.7%unknown effort	All 1 recorded result & sources 79.7% · raw 79.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MATH-Vision (Gemma,280 visual tokens)	76.7%unknown effort	All 1 recorded result & sources 76.7% · raw 76.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
MedXpertQA MM	48.7%unknown effort	All 1 recorded result & sources 48.7% · raw 48.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
MedXpertQA MM (Gemma,280 visual tokens)	47.4%unknown effort	All 1 recorded result & sources 47.4% · raw 47.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
InfographicVQA	88.4%unknown effort	All 1 recorded result & sources 88.4% · raw 88.4 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native %. https://arxiv.org/pdf/2607.02770#page=6
InfographicVQA (Gemma,280 visual tokens)	58.7%unknown effort	All 1 recorded result & sources 58.7% · raw 58.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
OmniDocBench1.5 (average edit distance)	0.164 edit distanceunknown effort	All 1 recorded result & sources 0.164 edit distance · raw 0.164 edit distance Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table6; thinking,1120 visual tokens; execution date and exact output budget unpublished. Native edit distance. https://arxiv.org/pdf/2607.02770#page=6
OmniDocBench1.5 (average edit distance) (Gemma,280 visual tokens)	0.408 edit distanceunknown effort	All 1 recorded result & sources 0.408 edit distance · raw 0.408 edit distance Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table12; thinking,280 visual tokens; separate visual-resolution configuration, not a new checkpoint. https://arxiv.org/pdf/2607.02770#page=17
RULER32K (Gemma,no thinking)	96.4%none effort	All 1 recorded result & sources 96.4% · raw 96.4 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. RULER32K. Native %. https://arxiv.org/pdf/2607.02770#page=8
RULER128K (Gemma,no thinking)	91.2%none effort	All 1 recorded result & sources 91.2% · raw 91.2 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. RULER128K. Native %. https://arxiv.org/pdf/2607.02770#page=8
LOFT text retrieval128K (Recall@k) (Gemma,no thinking)	66.4%none effort	All 1 recorded result & sources 66.4% · raw 66.4 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. LOFT text retrieval128K (Recall@k). Native %. https://arxiv.org/pdf/2607.02770#page=8
GraphWalks <128K (F1) (Gemma,no thinking)	71%none effort	All 1 recorded result & sources 71% · raw 71 % Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. GraphWalks <128K (F1). Native %. https://arxiv.org/pdf/2607.02770#page=8
MTOB English→Kalamang ~128K half-book (Gemma,no thinking)	45.1 chrFnone effort	All 1 recorded result & sources 45.1 chrF · raw 45.1 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB English→Kalamang ~128K half-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB English→Kalamang ~256K full-book (Gemma,no thinking)	41.9 chrFnone effort	All 1 recorded result & sources 41.9 chrF · raw 41.9 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB English→Kalamang ~256K full-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB Kalamang→English ~128K half-book (Gemma,no thinking)	37.3 chrFnone effort	All 1 recorded result & sources 37.3 chrF · raw 37.3 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB Kalamang→English ~128K half-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
MTOB Kalamang→English ~256K full-book (Gemma,no thinking)	32.9 chrFnone effort	All 1 recorded result & sources 32.9 chrF · raw 32.9 chrF Headline · none effort · Own vendor Source/record date: 2026-10-19 Harness: Google report Table9, thinking disabled. MTOB Kalamang→English ~256K full-book. Native chrF. https://arxiv.org/pdf/2607.02770#page=8
τ²-bench (Tau-2)	69%unknown effort	All 1 recorded result & sources 69% · raw 69 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google official card; published average over3 domains, thinking on. Not recomputed from report domains. https://huggingface.co/google/gemma-4-12B-it/blob/707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7/README.md
CoVoST (Gemma card aggregate)	38.50 BLEUunknown effort	All 1 recorded result & sources 38.50 BLEU · raw 38.5 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google card audio aggregate. CoVoST translation corpus BLEU.12B excludes Chinese. Distinct from report Table7/8. https://huggingface.co/google/gemma-4-12B-it/blob/707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7/README.md
FLEURS (Gemma card aggregate error)	6.90%unknown effort	All 1 recorded result & sources 6.90% · raw 6.9 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google card audio aggregate. FLEURS error fraction converted to percent; WER/CER language conventions are not restated on card.12B excludes Chinese. Distinct from report Table7/8. https://huggingface.co/google/gemma-4-12B-it/blob/707f0a3b8a3c7ad586ed01e27eafbad8a27dd0f7/README.md
CoVoST ja→English (Gemma report)	26.4 BLEUunknown effort	All 1 recorded result & sources 26.4 BLEU · raw 26.4 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; ja source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
CoVoST de→English (Gemma report)	41.9 BLEUunknown effort	All 1 recorded result & sources 41.9 BLEU · raw 41.9 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; de source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
CoVoST fr→English (Gemma report)	42.5 BLEUunknown effort	All 1 recorded result & sources 42.5 BLEU · raw 42.5 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; fr source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
CoVoST es→English (Gemma report)	44.6 BLEUunknown effort	All 1 recorded result & sources 44.6 BLEU · raw 44.6 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; es source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
CoVoST it→English (Gemma report)	43.3 BLEUunknown effort	All 1 recorded result & sources 43.3 BLEU · raw 43.3 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; it source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
CoVoST ru→English (Gemma report)	50.5 BLEUunknown effort	All 1 recorded result & sources 50.5 BLEU · raw 50.5 BLEU Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; ru source language; native corpus BLEU, not percent accuracy. https://arxiv.org/pdf/2607.02770#page=7
FLEURS en (WER,Gemma report)	6.30%unknown effort	All 1 recorded result & sources 6.30% · raw 6.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS ko (CER,Gemma report)	5.70%unknown effort	All 1 recorded result & sources 5.70% · raw 5.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription CER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS ja (CER,Gemma report)	8.00%unknown effort	All 1 recorded result & sources 8.00% · raw 8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription CER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS de (WER,Gemma report)	5.30%unknown effort	All 1 recorded result & sources 5.30% · raw 5.3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS fr (WER,Gemma report)	8.10%unknown effort	All 1 recorded result & sources 8.10% · raw 8.1 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS es (WER,Gemma report)	3.80%unknown effort	All 1 recorded result & sources 3.80% · raw 3.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS it (WER,Gemma report)	3.00%unknown effort	All 1 recorded result & sources 3.00% · raw 3 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS pt-br (WER,Gemma report)	4.70%unknown effort	All 1 recorded result & sources 4.70% · raw 4.7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS ru (WER,Gemma report)	6.80%unknown effort	All 1 recorded result & sources 6.80% · raw 6.8 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
FLEURS ar (WER,Gemma report)	7.00%unknown effort	All 1 recorded result & sources 7.00% · raw 7 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google report audio Table8; transcription WER. Source fraction retained. https://arxiv.org/pdf/2607.02770#page=7
DocVQA	94.9%unknown effort	All 1 recorded result & sources 94.9% · raw 94.9 % Headline · unknown effort · Own vendor Source/record date: 2026-10-19 Harness: Google12B launch chart, exact94.9 printed above12B bar. Peer bars lack labels and are not estimated from coordinates. https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

Independent evaluator headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
GDPval-AA v2.1 Elo · Artificial Analysis	427unknown effort	All 1 recorded result & sources 427 · raw 427 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-20 Manually observed public GDPval-AA v2.1 leaderboard table, full 298-configuration pool checked 20 October 2026. Exact source configuration: Gemma 4 12B (Reasoning); creator: Google; source rank 234; source release-month label Jun 2026 is model release metadata, not execution date. Published rounded native Elo 427; literal displayed CI -25 / +25. Native Elo is not a percentage and is not the normalized (Elo−500)/2000 chart quantity. This versioned metric is separate from older generic GDPval-AA Elo and vendor GDPval/GDPval-AA results; no old evidence is overwritten. Execution date, inference precision, unprinted effort budgets and task costs are unpublished in this captured table. Elo is relative to the dated comparison pool; drift from an earlier observation does not prove a new evaluation run. Source says Reasoning (on), not an API effort level; reasoning effort remains unknown. Headline: highest published native Elo for this exact approved model; tied values prefer existing featured effort then source order. https://artificialanalysis.ai/evaluations/gdpval-aa
Intelligence Index (estimated) · Artificial Analysis	14.2unknown effort	All 2 recorded results & sources 14.2 · raw 14.1847516647736 index Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 9.4 · raw 9.37989472421056 index Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
AA-Omniscience Index · Artificial Analysis	-52.7unknown effort	All 2 recorded results & sources -52.7 · raw -52.71666666666667 index Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b -52.8 · raw -52.8 index Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
AA-Omniscience accuracy · Artificial Analysis	15.6%unknown effort	All 2 recorded results & sources 15.6% · raw 15.616666666666667 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 11.9% · raw 11.883333333333333 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
AA-Omniscience hallucination rate · Artificial Analysis	81%unknown effort	All 2 recorded results & sources 81% · raw 80.979656330239 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 73.4% · raw 73.40646869680349 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
GDPval-AA Elo · Artificial Analysis	427.3unknown effort	All 1 recorded result & sources 427.3 · raw 427.31 Elo Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b
Terminal-Bench 2.1 (AA) · Artificial Analysis	27.3%unknown effort	All 1 recorded result & sources 27.3% · raw 27.340823970037498 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b
Terminal Bench Hard (AA) · Artificial Analysis	18.2%unknown effort	All 2 recorded results & sources 18.2% · raw 18.181818181818198 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 11.4% · raw 11.363636363636399 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
AA-LCR · Artificial Analysis	63.7%unknown effort	All 2 recorded results & sources 63.7% · raw 63.66666666666671 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 35% · raw 35 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
IFBench (AA) · Artificial Analysis	73.5%unknown effort	All 2 recorded results & sources 73.5% · raw 73.5374149659864 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 45.2% · raw 45.1700680272109 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
Humanity's Last Exam (AA harness) · Artificial Analysis	15.7%unknown effort	All 2 recorded results & sources 15.7% · raw 15.6626506024096 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 6.3% · raw 6.34847080630213 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
GPQA Diamond (AA) · Artificial Analysis	75.3%unknown effort	All 2 recorded results & sources 75.3% · raw 75.2525252525252 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 66.1% · raw 66.06060606060609 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
CritPt (AA) · Artificial Analysis	0%unknown effort	All 2 recorded results & sources 0% · raw 0 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 0% · raw 0 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
MMMU-Pro (AA) · Artificial Analysis	69.7%unknown effort	All 2 recorded results & sources 69.7% · raw 69.6531791907514 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 62% · raw 61.965317919075105 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
τ²-Telecom (AA) · Artificial Analysis	36.3%unknown effort	All 2 recorded results & sources 36.3% · raw 36.2573099415205 % Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 31.9% · raw 31.8713450292398 % Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning
Output speed · Artificial Analysis	117.5 tok/sunknown effort	All 2 recorded results & sources 117.5 tok/s · raw 117.479463907397 tok/s Headline · unknown effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Reasoning enabled without a published common API effort label. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b 121 tok/s · raw 121.02924258684 tok/s Alternative · none effort · Independent evaluator Source/record date: 2026-10-19 Independent AA exact model page explicitly links https://huggingface.co/google/gemma-4-12B-it. Explicit non-reasoning configuration. Index v4.3.2; inference precision, execution date and exact budget unpublished. Linking reference weights does not prove BF16 inference. Costs with unavailable/zero model prices are not imported as free tariffs. https://artificialanalysis.ai/models/gemma-4-12b-non-reasoning

Read how we select and source scores or the comparison guide.