Gemma 3 12B vs Phi 4 Mini Instruct
Wins 3 of 6 areas
Coding · Reasoning · Following instructions
Wins 1 of 6 areas
Long documents
Gemma 3 12B wins more areas, narrowly.Phi 4 Mini Instruct is better at long documents.
Scores updated · 23 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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.
- ReasoningHard problems that need careful thinking10Gemma 3 12B1 of 3 tests · 2 ties
- CodingWriting and fixing software10Gemma 3 12B1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked10Gemma 3 12B1 of 1 test
- Long documentsFinding answers in very long texts01Phi 4 Mini Instruct1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 2 ties
- FactsGetting facts right instead of making them up11Even1 each · 1 tie
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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Gemma 3 12B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+15.6points ahead
- Code for real scientific research problemsSciCode+5.6points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+1.8points ahead
Where Phi 4 Mini Instruct pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+19.4points ahead
- Reasons across sets of long documentsAA-LCR+7points ahead
Every test, side by side
All 23 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 3 12B
- SciCodeGemma 3 12B by 5.616.410.8+5.6
- Terminal-Bench Hardtie0.80tie
ReasoningGemma 3 12B
- GPQA DiamondGemma 3 12B by 1.834.933.1+1.8
- Humanity's Last Examtie4.24.5tie
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationPhi 4 Mini Instruct by 19.42.622+19.4
- Vectara HHEM hallucination ratelower is betterGemma 3 12B by 19.14.423.5+19.1
- AA-Omniscience · Accuracytie10.39.5tie
Following instructionsGemma 3 12B
- IFBenchGemma 3 12B by 15.636.721.1+15.6
Other results11 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.
- vectara_avg_summary_lengthPhi 4 Mini Instruct by 330.5 rating points89.7420.2+330.5 rating
- vectara_factual_consistencyGemma 3 12B by 19.195.676.5+19.1
- AA-OmnisciencePhi 4 Mini Instruct by 16-77.1-61.1+16
- BBHGemma 3 12B by 15.385.770.4+15.3
- Arena HardGemma 3 12B by 10.843.632.8+10.8
- MMLU-ProGemma 3 12B by 7.860.652.8+7.8
- GSM8KGemma 3 12B by 5.894.488.6+5.8
- vectara_answer_rateGemma 3 12B by 4.997.492.5+4.9
- τ²-Bench Telecom (AA run)Gemma 3 12B by 2.610.88.2+2.6
- AA IntelligencePhi 4 Mini Instruct by 2.53.86.3+2.5
- Artificial Analysis Coding IndexGemma 3 12B by 25.83.8+2
Questions people ask
Which is better, Gemma 3 12B or Phi 4 Mini Instruct?
Gemma 3 12B wins three of the six areas where both have results: coding, reasoning and following instructions. Phi 4 Mini Instruct wins long documents. They are level on agents and facts.
Which is better for coding?
Gemma 3 12B. It wins 1 of the 2 coding tests both models report; Phi 4 Mini Instruct wins none, and 1 is a tie.
How do you compare the two?
We use the 23 benchmark tests both models have published scores on. The verdict counts the 12 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 11 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.