Gemma 3 27B vs Gemma 4 31B
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Gemma 4 31B is the stronger all-rounder.Gemma 3 27B is cheaper.
Scores updated · 37 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. 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 software05Gemma 4 31B5 of 5 tests
- Images and chartsUnderstanding pictures, charts and video04Gemma 4 31B4 of 4 tests
- ReasoningHard problems that need careful thinking03Gemma 4 31B3 of 3 tests
- AgentsCarrying out multi-step tasks on its own02Gemma 4 31B2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up02Gemma 4 31B2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts02Gemma 4 31B2 of 2 tests
- MathCompetition and research-level math01Gemma 4 31B1 of 1 test
- Following instructionsDoing exactly what it is asked01Gemma 4 31B1 of 1 test
Math 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 58% 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 4 31B pulls ahead
- US invitational high-school math exam problemsAIME 2026+68.4points ahead
- Reasons across sets of long documentsAA-LCR+62.4points ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+52.9points ahead
Where Gemma 3 27B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 37 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 4 31B
- LiveCodeBench v6Gemma 4 31B by 50.929.180+50.9
- Terminal-Bench 2.1Gemma 4 31B by 38.94.543.4+38.9
- Terminal-Bench HardGemma 4 31B by 32.63.836.4+32.6
- SWE-bench ProGemma 4 31B by 24.311.435.7+24.3
- SciCodeGemma 4 31B by 22.223.345.5+22.2
AgentsGemma 4 31B
- τ-Bench V3 · BankingGemma 4 31B by 140.814.8+14
- GDPValGemma 4 31B by 6.106.1+6.1
- Terminal-Bench 4.0tie00tie
ReasoningGemma 4 31B
- GPQA DiamondGemma 4 31B by 42.942.885.7+42.9
- Humanity's Last ExamGemma 4 31B by 19.24.423.6+19.2
- CritPtGemma 4 31B by 1.401.4+1.4
FactsGemma 4 31B
- AA-Omniscience · Non-hallucinationGemma 4 31B by 7.17.915+7.1
- AA-Omniscience · AccuracyGemma 4 31B by 712.920+7
- Vectara HHEM hallucination ratelower is bettertie7.47.4tie
Images and chartsGemma 4 31B
- MMMU-ProGemma 4 31B by 25.44873.4+25.4
- MMMUGemma 4 31B by 15.564.980.4+15.5
- MathVistaGemma 4 31B by 11.767.679.3+11.7
- LMArena · VisionGemma 4 31B by 113 rating points11641277+113 rating
Long documentsGemma 4 31B
- AA-LCRGemma 4 31B by 62.47.369.7+62.4
- MRCR v2 8 needle 128k (average)Gemma 4 31B by 52.913.566.4+52.9
Following instructionsGemma 4 31B
- IFBenchGemma 4 31B by 43.831.875.6+43.8
Other results15 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.
- Codeforces ELOGemma 4 31B by 2040 rating points1102150+2040 rating
- τ²-BenchGemma 4 31B by 60.716.276.9+60.7
- BIG-Bench Extra HardGemma 4 31B by 55.119.374.4+55.1
- τ²-Bench Telecom (AA run)Gemma 4 31B by 49.410.559.9+49.4
- MATH-VisionGemma 4 31B by 39.64685.6+39.6
- Artificial Analysis Coding IndexGemma 4 31B by 33.310.143.4+33.3
- vectara_avg_summary_lengthGemma 3 27B by 20.696.475.8+20.6
- AA-OmniscienceGemma 4 31B by 19.3-67.2-47.9+19.3
- MMMLUGemma 4 31B by 17.770.788.4+17.7
- MMLU-ProGemma 4 31B by 17.667.685.2+17.6
- AA IntelligenceGemma 4 31B by 9.84.914.7+9.8
- AI2DGemma 4 31B by 4.584.589+4.5
- vectara_answer_rateGemma 4 31B by 1.298.8100+1.2
- OmniDocBench 1.5 (average edit distance, lower is better)lower is bettertie0.40.1tie
- vectara_factual_consistencytie92.692.6tie
Questions people ask
Which is better, Gemma 3 27B or Gemma 4 31B?
Gemma 4 31B wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Gemma 3 27B wins none, but costs 58% less.
Which is better for coding?
Gemma 4 31B. It wins 5 of the 5 coding tests both models report; Gemma 3 27B wins none.
Which is cheaper?
Gemma 3 27B costs $0.08 per million input tokens and $0.16 per million output tokens; Gemma 4 31B costs $0.17 and $0.40. That makes Gemma 3 27B about 58% cheaper for the same work.
How do you compare the two?
We use the 37 benchmark tests both models have published scores on. The verdict counts the 22 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 15 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.