Gemma 3 27B vs Phi 4
Wins 2 of 5 areas
Long documents · Following instructions
Wins 3 of 5 areas
Coding · Reasoning · Facts
Phi 4 is the stronger all-rounder.Gemma 3 27B is cheaper and better at long documents.
Scores updated · 24 tests both models report · How we compare
Where each one wins
Tests won in each of the five 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.
- FactsGetting facts right instead of making them up03Phi 43 of 3 tests
- ReasoningHard problems that need careful thinking01Phi 41 of 3 tests · 2 ties
- CodingWriting and fixing software01Phi 41 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10Gemma 3 27B1 of 1 test
- Following instructionsDoing exactly what it is asked10Gemma 3 27B1 of 1 test
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 27B costs 62% 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 Phi 4 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+14.7points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+10.9points ahead
- Code for real scientific research problemsSciCode+2.7points ahead
Where Gemma 3 27B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+8.3points ahead
Every test, side by side
All 24 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningPhi 4
- GPQA DiamondPhi 4 by 14.742.857.5+14.7
- Humanity's Last Examtie4.43.8tie
- CritPttie00tie
FactsPhi 4
- AA-Omniscience · Non-hallucinationPhi 4 by 10.97.918.8+10.9
- Vectara HHEM hallucination ratelower is betterPhi 4 by 3.77.43.7+3.7
- AA-Omniscience · AccuracyPhi 4 by 1.112.914.1+1.1
Following instructionsGemma 3 27B
- IFBenchGemma 3 27B by 8.331.823.5+8.3
Other results14 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.
- IFEvalGemma 3 27B by 27.490.463+27.4
- vectara_answer_rateGemma 3 27B by 18.198.880.7+18.1
- AA-OmnisciencePhi 4 by 11.5-67.2-55.7+11.5
- τ²-Bench Telecom (AA run)Gemma 3 27B by 10.510.50+10.5
- MATHGemma 3 27B by 8.68980.4+8.6
- MMLUPhi 4 by 7.976.984.8+7.9
- SimpleQAGemma 3 27B by 7103+7
- MMLU-ProXGemma 3 27B by 6.756.649.9+6.7
- HumanEvalGemma 3 27B by 5.287.882.6+5.2
- vectara_factual_consistencyPhi 4 by 3.792.696.3+3.7
- MMLU-ProPhi 4 by 2.867.670.4+2.8
- vectara_avg_summary_lengthPhi 4 by 24.5 rating points96.4120.9+24.5 rating
- Artificial Analysis Coding IndexPhi 4 by 1.110.111.2+1.1
- AA IntelligencePhi 4 by 14.95.9+1
Questions people ask
Which is better, Gemma 3 27B or Phi 4?
Phi 4 wins three of the five areas where both have results: coding, reasoning and facts. Gemma 3 27B wins long documents and following instructions, and costs 62% less.
Which is better for coding?
Phi 4. It wins 1 of the 2 coding tests both models report; Gemma 3 27B wins none, and 1 is a tie.
Which is cheaper?
Gemma 3 27B costs $0.08 per million input tokens and $0.16 per million output tokens; Phi 4 costs $0.13 and $0.50. That makes Gemma 3 27B about 62% cheaper for the same work.
How do you compare the two?
We use the 24 benchmark tests both models have published scores on. The verdict counts the 10 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 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.