Gemma 3 4B vs Nova Lite
Wins 0 of 7 areas
—
Wins 5 of 7 areas
Coding · Facts · Images and charts · Long documents · Following instructions
Nova Lite 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 up02Nova Lite2 of 3 tests · 1 tie
- CodingWriting and fixing software01Nova Lite1 of 2 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video01Nova Lite1 of 1 test
- Long documentsFinding answers in very long texts01Nova Lite1 of 1 test
- Following instructionsDoing exactly what it is asked01Nova Lite1 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 50% 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 Lite pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+40.6points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+14.2points ahead
- Reasons across sets of long documentsAA-LCR+12.3points ahead
Where Gemma 3 4B pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+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.
ReasoningEven
- GPQA DiamondNova Lite by 14.229.143.3+14.2
- Humanity's Last ExamGemma 3 4B by 15.34.3+1
- CritPttie00tie
FactsNova Lite
- AA-Omniscience · Non-hallucinationNova Lite by 40.61.842.4+40.6
- AA-Omniscience · AccuracyNova Lite by 2.17.79.8+2.1
- Vectara HHEM hallucination ratelower is bettertie6.46.1tie
Following instructionsNova Lite
- IFBenchNova Lite by 5.828.334.1+5.8
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 Lite by 40.7-82.9-42.2+40.7
- vectara_answer_rateNova Lite by 32.667.399.9+32.6
- TextVQANova Lite by 22.457.880.2+22.4
- ChartQANova Lite by 1868.886.8+18
- DocVQANova Lite by 16.675.892.4+16.6
- vectara_avg_summary_lengthNova Lite by 14.477.491.8+14.4
- HumanEvalNova Lite by 14.171.385.4+14.1
- τ²-Bench Telecom (AA run)Nova Lite by 12.5517.5+12.5
- MMLUNova Lite by 10.958.469.3+10.9
- BBHNova Lite by 10.272.282.4+10.2
- GSM8KNova Lite by 5.389.294.5+5.3
- AA Agentic IndexNova Lite by 4.11.75.8+4.1
- Artificial Analysis Coding IndexNova Lite by 2.42.75.1+2.4
- AA IntelligenceNova Lite by 1.94.86.7+1.9
- IFEvaltie90.289.7tie
- vectara_factual_consistencytie93.693.9tie
Questions people ask
Which is better, Gemma 3 4B or Nova Lite?
Nova Lite 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 50% less. They are level on agents and reasoning.
Which is better for coding?
Nova Lite. It wins 1 of the 2 coding tests both models report; Gemma 3 4B wins none, and 1 is a tie.
Which is cheaper?
Gemma 3 4B costs $0.05 per million input tokens and $0.10 per million output tokens; Nova Lite costs $0.06 and $0.24. That makes Gemma 3 4B about 50% 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.