Nova Lite vs Qwen3.5 35B A3B
Wins 1 of 7 areas
Facts
Wins 6 of 7 areas
Coding · Agents · Reasoning · Images and charts · Long documents · Following instructions
Qwen3.5 35B A3B is the stronger all-rounder.Nova Lite is cheaper and better at facts.
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.
- ReasoningHard problems that need careful thinking02Qwen3.5 35B A3B2 of 3 tests · 1 tie
- CodingWriting and fixing software02Qwen3.5 35B A3B2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video02Qwen3.5 35B A3B2 of 2 tests
- AgentsCarrying out multi-step tasks on its own01Qwen3.5 35B A3B1 of 1 test
- Long documentsFinding answers in very long texts01Qwen3.5 35B A3B1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3.5 35B A3B1 of 1 test
- FactsGetting facts right instead of making them up21Nova Lite2 of 3 tests
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.
Nova Lite costs 87% 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 Qwen3.5 35B A3B pulls ahead
- Reasons across sets of long documentsAA-LCR+53points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+41.2points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+38.4points ahead
Where Nova Lite pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+27.8points 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.
CodingQwen3.5 35B A3B
- Terminal-Bench HardQwen3.5 35B A3B by 25.70.826.5+25.7
- SciCodeQwen3.5 35B A3B by 23.813.937.7+23.8
ReasoningQwen3.5 35B A3B
- GPQA DiamondQwen3.5 35B A3B by 41.243.384.5+41.2
- Humanity's Last ExamQwen3.5 35B A3B by 16.74.321+16.7
- CritPttie00.9tie
FactsNova Lite
- AA-Omniscience · Non-hallucinationNova Lite by 27.842.414.6+27.8
- AA-Omniscience · AccuracyQwen3.5 35B A3B by 10.39.820.1+10.3
- Vectara HHEM hallucination ratelower is betterNova Lite by 4.46.110.5+4.4
Images and chartsQwen3.5 35B A3B
Following instructionsQwen3.5 35B A3B
- IFBenchQwen3.5 35B A3B by 38.434.172.5+38.4
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.
- τ²-Bench Telecom (AA run)Qwen3.5 35B A3B by 71.717.589.2+71.7
- OpenBookQANova Lite by 48.692.844.2+48.6
- Artificial Analysis Coding IndexQwen3.5 35B A3B by 31.95.137+31.9
- LVBenchQwen3.5 35B A3B by 3140.471.4+31
- HumanEvalNova Lite by 18.985.466.5+18.9
- AA IntelligenceQwen3.5 35B A3B by 12.66.719.3+12.6
- MMLUQwen3.5 35B A3B by 11.869.381.1+11.8
- AA Agentic IndexQwen3.5 35B A3B by 65.811.8+6
- AA-OmniscienceNova Lite by 5.9-42.2-48.1+5.9
- vectara_factual_consistencyNova Lite by 4.493.989.5+4.4
- vectara_avg_summary_lengthQwen3.5 35B A3B by 3.191.894.9+3.1
- ARC-ChallengeQwen3.5 35B A3B by 392.495.4+3
- IFEvalQwen3.5 35B A3B by 2.289.791.9+2.2
- BFCLv4tie66.667.3tie
- vectara_answer_ratetie99.999.8tie
Questions people ask
Which is better, Nova Lite or Qwen3.5 35B A3B?
Qwen3.5 35B A3B wins six of the seven areas where both have results: coding, agents, reasoning, images and charts, long documents and following instructions. Nova Lite wins facts, and costs 87% less.
Which is better for coding?
Qwen3.5 35B A3B. It wins 2 of the 2 coding tests both models report; Nova Lite wins none.
Which is cheaper?
Nova Lite costs $0.06 per million input tokens and $0.24 per million output tokens; Qwen3.5 35B A3B costs $0.25 and $2.00. That makes Nova Lite about 87% 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 13 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.