Gemma 4 E2B vs Qwen3.5 2B
Wins 3 of 7 areas
Coding · Agents · Following instructions
Wins 2 of 7 areas
Images and charts · Long documents
Gemma 4 E2B wins more areas, narrowly.Qwen3.5 2B is better at images and charts.
Scores updated · 58 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 software10Gemma 4 E2B1 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own10Gemma 4 E2B1 of 2 tests · 1 tie
- Following instructionsDoing exactly what it is asked10Gemma 4 E2B1 of 1 test
- Images and chartsUnderstanding pictures, charts and video14Qwen3.5 2B4 of 5 tests
- Long documentsFinding answers in very long texts01Qwen3.5 2B1 of 1 test
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
Long documents and following instructions rest on a single test each.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemma 4 E2B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+41.8points ahead
- Code for real scientific research problemsSciCode+18.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+6.5points ahead
Every test, side by side
All 58 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 4 E2B
- SciCodeGemma 4 E2B by 18.120.92.8+18.1
- Terminal-Bench Hardtie33.8tie
ReasoningEven
- GPQA DiamondQwen3.5 2B by 2.343.345.6+2.3
- Humanity's Last ExamGemma 4 E2B by 2.24.82.6+2.2
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationGemma 4 E2B by 41.867.625.8+41.8
- AA-Omniscience · AccuracyQwen3.5 2B by 1.56.68.1+1.5
Images and chartsQwen3.5 2B
Following instructionsGemma 4 E2B
- IFBenchGemma 4 E2B by 6.53831.5+6.5
Other results42 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.
- τ²-Bench Telecom (AA run)Qwen3.5 2B by 48.220.869+48.2
- ScreenSpot-v2 WebQwen3.5 2B by 43.522.465.9+43.5
- AA-OmniscienceGemma 4 E2B by 36.6-23.6-60.2+36.6
- ScreenSpot-v2 DesktopQwen3.5 2B by 35.728.163.8+35.7
- ChartQA TestQwen3.5 2B by 35.243.278.4+35.2
- ScreenSpot-v2 MobileQwen3.5 2B by 26.842.969.7+26.8
- MMEQwen3.5 2B by 21.354.976.2+21.3
- t2-benchQwen3.5 2B by 19.429.448.8+19.4
- InfographicVQA (val)Qwen3.5 2B by 19.154.473.5+19.1
- GQA TestDev_BalancedQwen3.5 2B by 15.238.753.9+15.2
- MMLU testGemma 4 E2B by 14.969.254.3+14.9
- TextVQA-valQwen3.5 2B by 14.862.577.3+14.8
- MMLU-Pro testGemma 4 E2B by 14.344.129.8+14.3
- OCRBench v1Qwen3.5 2B by 14.270.284.4+14.2
- PerceptionQwen3.5 2B by 12.159.171.2+12.1
- MM-IFEvalGemma 4 E2B by 10.265.655.4+10.2
- Math & logical reasoningQwen3.5 2B by 9.251.861+9.2
- MMBenchQwen3.5 2B by 8.964.273.1+8.9
- SimpleVQAQwen3.5 2B by 7.927.335.2+7.9
- GeneralQwen3.5 2B by 7.156.964+7.1
- Multilingual MMBenchQwen3.5 2B by 7.162.869.9+7.1
- DocVQA-valQwen3.5 2B by 6.985.792.6+6.9
- Document, OCR & ChartQwen3.5 2B by 6.872.379.1+6.8
- MMBench DEV_EN_V11Qwen3.5 2B by 6.769.376+6.7
- BFCLv4Qwen3.5 2B by 6.63743.6+6.6
- MMLU-ProQwen3.5 2B by 6.56066.5+6.5
- HallusionBenchQwen3.5 2B by 5.759.865.5+5.7
- VLM Judge ScoreQwen3.5 2B by 5.384.489.7+5.3
- RealWorldQAQwen3.5 2B by 5.16065.1+5.1
- OCRBench v2_enQwen3.5 2B by 4.643.548.1+4.6
- POPEQwen3.5 2B by 4.68488.6+4.6
- IFEvalGemma 4 E2B by 4.48378.6+4.4
- SEED-Bench (image)Qwen3.5 2B by 4.471.475.8+4.4
- Artificial Analysis Coding IndexGemma 4 E2B by 4.37.22.9+4.3
- MMMLUGemma 4 E2B by 4.367.463.1+4.3
- MMMU (val) (Pass@1)Qwen3.5 2B by 341.144.1+3
- Multilingual MMMBQwen3.5 2B by 2.673.375.9+2.6
- MTL MMBench_DEVQwen3.5 2B by 2.164.866.9+2.1
- AA Intelligencetie7.86.9tie
- MMMU DEV_VALtie47.747.4tie
- MMMBtie74.374.5tie
- MMBench_DEVtie0.60.7tie
Questions people ask
Which is better, Gemma 4 E2B or Qwen3.5 2B?
Gemma 4 E2B wins three of the seven areas where both have results: coding, agents and following instructions. Qwen3.5 2B wins images and charts and long documents. They are level on reasoning and facts.
Which is better for coding?
Gemma 4 E2B. It wins 1 of the 2 coding tests both models report; Qwen3.5 2B wins none, and 1 is a tie.
How do you compare the two?
We use the 58 benchmark tests both models have published scores on. The verdict counts the 16 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 42 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.