Qwen3 235B A22B vs Qwen3 32B
Wins 5 of 6 areas
Coding · Agents · Reasoning · Facts · Following instructions
Wins 0 of 6 areas
—
Qwen3 235B A22B is the stronger all-rounder.Qwen3 32B is cheaper.
Scores updated · 26 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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 thinking20Qwen3 235B A22B2 of 3 tests · 1 tie
- CodingWriting and fixing software20Qwen3 235B A22B2 of 2 tests
- FactsGetting facts right instead of making them up21Qwen3 235B A22B2 of 3 tests
- AgentsCarrying out multi-step tasks on its own10Qwen3 235B A22B1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 235B A22B1 of 1 test
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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.
Qwen3 32B costs 77% 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 235B A22B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+4.5points ahead
- Code for real scientific research problemsSciCode+3.9points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+3.6points ahead
Where Qwen3 32B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 26 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3 235B A22B
- SciCodeQwen3 235B A22B by 3.939.936+3.9
- Terminal-Bench HardQwen3 235B A22B by 3.16.13+3.1
ReasoningQwen3 235B A22B
- Humanity's Last ExamQwen3 235B A22B by 3.6117.4+3.6
- GPQA DiamondQwen3 235B A22B by 3.27066.8+3.2
- CritPttie00.3tie
FactsQwen3 235B A22B
- AA-Omniscience · Non-hallucinationQwen3 235B A22B by 4.522.417.9+4.5
- Vectara HHEM hallucination ratelower is betterQwen3 32B by 3.49.35.9+3.4
- AA-Omniscience · AccuracyQwen3 235B A22B by 1.118.517.4+1.1
Following instructionsQwen3 235B A22B
- IFBenchQwen3 235B A22B by 2.438.736.3+2.4
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.
- MMLU-ProQwen3 32B by 13.668.281.8+13.6
- AIME 2025Qwen3 235B A22B by 8.681.572.9+8.6
- MMLU-ProXQwen3 235B A22B by 6.866.759.9+6.8
- τ²-Bench Telecom (AA run)Qwen3 32B by 5.82429.8+5.8
- AA-OmniscienceQwen3 235B A22B by 5.7-44.7-50.4+5.7
- LiveCodeBenchQwen3 235B A22B by 570.765.7+5
- vectara_answer_rateQwen3 32B by 594.999.9+5
- AIME 2024Qwen3 235B A22B by 4.385.781.4+4.3
- vectara_factual_consistencyQwen3 32B by 3.490.794.1+3.4
- LiveBenchQwen3 235B A22B by 2.277.174.9+2.2
- Artificial Analysis Coding IndexQwen3 235B A22B by 2.117.415.3+2.1
- Arena HardQwen3 235B A22B by 1.895.693.8+1.8
- vectara_avg_summary_lengthQwen3 32B by 10.2 rating points105.6115.8+10.2 rating
- AA Intelligencetie9.58.6tie
- BFCLv4tie70.870.3tie
Questions people ask
Which is better, Qwen3 235B A22B or Qwen3 32B?
Qwen3 235B A22B wins five of the six areas where both have results: coding, agents, reasoning, facts and following instructions. Qwen3 32B wins none, but costs 77% less. They are level on long documents.
Which is better for coding?
Qwen3 235B A22B. It wins 2 of the 2 coding tests both models report; Qwen3 32B wins none.
Which is cheaper?
Qwen3 235B A22B costs $0.70 per million input tokens and $2.80 per million output tokens; Qwen3 32B costs $0.16 and $0.64. That makes Qwen3 32B about 77% cheaper for the same work.
How do you compare the two?
We use the 26 benchmark tests both models have published scores on. The verdict counts the 11 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.