Gemma 3 4B vs Nova Pro
Wins 0 of 7 areas
—
Wins 5 of 7 areas
Coding · Facts · Images and charts · Long documents · Following instructions
Nova Pro is the stronger all-rounder.Gemma 3 4B is cheaper.
Scores updated · 28 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.
- FactsGetting facts right instead of making them up03Nova Pro3 of 3 tests
- CodingWriting and fixing software02Nova Pro2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video01Nova Pro1 of 1 test
- Long documentsFinding answers in very long texts01Nova Pro1 of 1 test
- Following instructionsDoing exactly what it is asked01Nova Pro1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- ReasoningHard problems that need careful thinking11Even1 each · 1 tie
Agents, images and charts, 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 4B costs 96% 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 Nova Pro pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+20.8points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+20.5points ahead
- Harder college exam questions with imagesMMMU-Pro+14.4points ahead
Where Gemma 3 4B pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+2.1points ahead
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingNova Pro
- SciCodeNova Pro by 13.57.320.8+13.5
- Terminal-Bench HardNova Pro by 5.30.86.1+5.3
ReasoningEven
- GPQA DiamondNova Pro by 20.829.149.9+20.8
- Humanity's Last ExamGemma 3 4B by 2.15.33.2+2.1
- CritPttie00tie
FactsNova Pro
- AA-Omniscience · Non-hallucinationNova Pro by 20.51.822.3+20.5
- AA-Omniscience · AccuracyNova Pro by 9.27.716.9+9.2
- Vectara HHEM hallucination ratelower is betterNova Pro by 1.36.45.1+1.3
Other results16 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.
- AA-OmniscienceNova Pro by 35.2-82.9-47.7+35.2
- vectara_answer_rateNova Pro by 3267.399.3+32
- TextVQANova Pro by 23.757.881.5+23.7
- ChartQANova Pro by 20.468.889.2+20.4
- DocVQANova Pro by 17.775.893.5+17.7
- HumanEvalNova Pro by 17.771.389+17.7
- MMLUNova Pro by 17.458.475.8+17.4
- BBHNova Pro by 14.772.286.9+14.7
- vectara_avg_summary_lengthGemma 3 4B by 11.277.466.2+11.2
- τ²-Bench Telecom (AA run)Nova Pro by 9514+9
- Artificial Analysis Coding IndexNova Pro by 8.32.711+8.3
- GSM8KNova Pro by 5.689.294.8+5.6
- AA Agentic IndexNova Pro by 31.74.7+3
- AA IntelligenceNova Pro by 2.24.87+2.2
- IFEvalNova Pro by 1.990.292.1+1.9
- vectara_factual_consistencyNova Pro by 1.393.694.9+1.3
Questions people ask
Which is better, Gemma 3 4B or Nova Pro?
Nova Pro wins five of the seven areas where both have results: coding, facts, images and charts, long documents and following instructions. Gemma 3 4B wins none, but costs 96% less. They are level on agents and reasoning.
Which is better for coding?
Nova Pro. It wins 2 of the 2 coding tests both models report; Gemma 3 4B wins none.
Which is cheaper?
Gemma 3 4B costs $0.05 per million input tokens and $0.10 per million output tokens; Nova Pro costs $0.80 and $3.20. That makes Gemma 3 4B about 96% cheaper for the same work.
How do you compare the two?
We use the 28 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 16 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.