Gemma 3 12B vs Gemma 3 27B
Wins 2 of 7 areas
Long documents · Following instructions
Wins 4 of 7 areas
Coding · Reasoning · Facts · Images and charts
Gemma 3 27B wins more areas, narrowly.Gemma 3 12B is cheaper and better at long documents.
Scores updated · 47 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software03Gemma 3 27B3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03Gemma 3 27B3 of 3 tests
- ReasoningHard problems that need careful thinking01Gemma 3 27B1 of 3 tests · 2 ties
- FactsGetting facts right instead of making them up12Gemma 3 27B2 of 3 tests
- Long documentsFinding answers in very long texts10Gemma 3 12B1 of 1 test
- Following instructionsDoing exactly what it is asked10Gemma 3 12B1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 3 ties
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Gemma 3 12B costs 17% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Gemma 3 27B pulls ahead
- Harder college exam questions with imagesMMMU-Pro+10.5points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+7.9points ahead
- Code for real scientific research problemsSciCode+6.9points ahead
Every test, side by side
All 47 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 3 27B
- SciCodeGemma 3 27B by 6.916.423.3+6.9
- Terminal-Bench 2.1Gemma 3 27B by 4.504.5+4.5
- Terminal-Bench HardGemma 3 27B by 30.83.8+3
AgentsEven
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
- τ-Bench V3 · Bankingtie0.80.8tie
ReasoningGemma 3 27B
- GPQA DiamondGemma 3 27B by 7.934.942.8+7.9
- Humanity's Last Examtie4.24.4tie
- CritPttie00tie
FactsGemma 3 27B
- AA-Omniscience · Non-hallucinationGemma 3 27B by 5.32.67.9+5.3
- Vectara HHEM hallucination ratelower is betterGemma 3 12B by 34.47.4+3
- AA-Omniscience · AccuracyGemma 3 27B by 2.710.312.9+2.7
Images and chartsGemma 3 27B
Following instructionsGemma 3 12B
- IFBenchGemma 3 12B by 4.936.731.8+4.9
Other results30 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- MATH-VisionGemma 3 27B by 13.932.146+13.9
- AA-OmniscienceGemma 3 27B by 9.9-77.1-67.2+9.9
- MMLU-ProGemma 3 27B by 760.667.6+7
- MMStar (Acc)Gemma 3 27B by 756.163.1+7
- vectara_avg_summary_lengthGemma 3 27B by 6.789.796.4+6.7
- MMVet (Acc)Gemma 3 27B by 6.164.971+6.1
- InfoVQAGemma 3 27B by 5.764.970.6+5.7
- Global-MMLU-LiteGemma 3 27B by 5.669.575.1+5.6
- MMMU (val) (Pass@1)Gemma 3 27B by 5.359.664.9+5.3
- LiveCodeBenchGemma 3 27B by 5.124.629.7+5.1
- OCRBench (Acc)Gemma 3 27B by 51 rating points702753+51 rating
- Artificial Analysis Coding IndexGemma 3 27B by 4.35.810.1+4.3
- MMBench-EN-v1.1 (Acc)Gemma 3 27B by 4.374.678.9+4.3
- SimpleQAGemma 3 27B by 3.76.310+3.7
- MathVision (full) (Pass@1)Gemma 3 27B by 3.432.135.5+3.4
- RealWorldQA (Acc)Gemma 3 27B by 3.459.162.5+3.4
- BIG-Bench Extra HardGemma 3 27B by 316.319.3+3
- vectara_factual_consistencyGemma 3 12B by 395.692.6+3
- TextVQAGemma 3 12B by 2.667.765.1+2.6
- HumanEvalGemma 3 27B by 2.485.487.8+2.4
- ChartQAGemma 3 27B by 2.375.778+2.3
- BBHGemma 3 27B by 1.985.787.6+1.9
- GSM8KGemma 3 27B by 1.594.495.9+1.5
- IFEvalGemma 3 27B by 1.588.990.4+1.5
- MBPPGemma 3 27B by 1.47374.4+1.4
- vectara_answer_rateGemma 3 27B by 1.497.498.8+1.4
- AA IntelligenceGemma 3 27B by 1.13.84.9+1.1
- DocVQAtie87.186.6tie
- AI2Dtie84.284.5tie
- τ²-Bench Telecom (AA run)tie10.810.5tie
Questions people ask
Which is better, Gemma 3 12B or Gemma 3 27B?
Gemma 3 27B wins four of the seven areas where both have results: coding, reasoning, facts and images and charts. Gemma 3 12B wins long documents and following instructions, and costs 17% less. They are level on agents.
Which is better for coding?
Gemma 3 27B. It wins 3 of the 3 coding tests both models report; Gemma 3 12B wins none.
Which is cheaper?
Gemma 3 12B costs $0.05 per million input tokens and $0.15 per million output tokens; Gemma 3 27B costs $0.08 and $0.16. That makes Gemma 3 12B about 17% cheaper for the same work.
How do you compare the two?
We use the 47 benchmark tests both models have published scores on. The verdict counts the 17 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 30 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.