Gemma 3 27B vs Nova Lite
Wins 2 of 7 areas
Coding · Images and charts
Wins 3 of 7 areas
Facts · Long documents · Following instructions
Nova Lite wins more areas, narrowly.Gemma 3 27B is cheaper and better at coding.
Scores updated · 29 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 up12Nova Lite2 of 3 tests
- Long documentsFinding answers in very long texts01Nova Lite1 of 1 test
- Following instructionsDoing exactly what it is asked01Nova Lite1 of 1 test
- CodingWriting and fixing software20Gemma 3 27B2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video20Gemma 3 27B2 of 2 tests
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- ReasoningHard problems that need careful thinking00Even0 each · 3 ties
Agents, 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 20% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Nova Lite pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+34.5points ahead
- Reasons across sets of long documentsAA-LCR+11.7points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+2.3points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemma 3 27B
- SciCodeGemma 3 27B by 9.423.313.9+9.4
- Terminal-Bench HardGemma 3 27B by 33.80.8+3
ReasoningEven
- GPQA Diamondtie42.843.3tie
- Humanity's Last Examtie4.44.3tie
- CritPttie00tie
FactsNova Lite
- AA-Omniscience · Non-hallucinationNova Lite by 34.57.942.4+34.5
- AA-Omniscience · AccuracyGemma 3 27B by 3.212.99.8+3.2
- Vectara HHEM hallucination ratelower is betterNova Lite by 1.37.46.1+1.3
Images and chartsGemma 3 27B
Following instructionsNova Lite
- IFBenchNova Lite by 2.331.834.1+2.3
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 Lite by 25-67.2-42.2+25
- TextVQANova Lite by 15.165.180.2+15.1
- MATHGemma 3 27B by 11.18977.9+11.1
- ChartQANova Lite by 8.87886.8+8.8
- MMLUGemma 3 27B by 7.676.969.3+7.6
- τ²-Bench Telecom (AA run)Nova Lite by 710.517.5+7
- DocVQANova Lite by 5.886.692.4+5.8
- BBHGemma 3 27B by 5.287.682.4+5.2
- Artificial Analysis Coding IndexGemma 3 27B by 510.15.1+5
- vectara_avg_summary_lengthGemma 3 27B by 4.696.491.8+4.6
- HumanEvalGemma 3 27B by 2.487.885.4+2.4
- AA IntelligenceNova Lite by 1.84.96.7+1.8
- GSM8KGemma 3 27B by 1.495.994.5+1.4
- vectara_factual_consistencyNova Lite by 1.392.693.9+1.3
- vectara_answer_rateNova Lite by 1.198.899.9+1.1
- IFEvaltie90.489.7tie
Questions people ask
Which is better, Gemma 3 27B or Nova Lite?
Nova Lite wins three of the seven areas where both have results: facts, long documents and following instructions. Gemma 3 27B wins coding and images and charts, and costs 20% less. They are level on agents and reasoning.
Which is better for coding?
Gemma 3 27B. It wins 2 of the 2 coding tests both models report; Nova Lite wins none.
Which is cheaper?
Gemma 3 27B costs $0.08 per million input tokens and $0.16 per million output tokens; Nova Lite costs $0.06 and $0.24. That makes Gemma 3 27B about 20% cheaper for the same work.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 13 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.