GPT-5.2 vs Qwen3.5 397B A17B
Wins 4 of 8 areas
Coding · Reasoning · Facts · Math
Wins 2 of 8 areas
Images and charts · Following instructions
GPT-5.2 wins more areas, narrowly.Qwen3.5 397B A17B is cheaper and better at images and charts.
Scores updated · 40 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- MathCompetition and research-level math40GPT-5.24 of 4 tests
- ReasoningHard problems that need careful thinking30GPT-5.23 of 3 tests
- FactsGetting facts right instead of making them up30GPT-5.23 of 3 tests
- CodingWriting and fixing software42GPT-5.24 of 6 tests
- Images and chartsUnderstanding pictures, charts and video02Qwen3.5 397B A17B2 of 2 tests
- Following instructionsDoing exactly what it is asked01Qwen3.5 397B A17B1 of 1 test
- AgentsCarrying out multi-step tasks on its own22Even2 each
- Long documentsFinding answers in very long texts11Even1 each
Following instructions rests on a single test.
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.
Qwen3.5 397B A17B costs 73% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.2 pulls ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+36.2points ahead
- Real work tasks from 44 professionsGDPVal+33.5points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+13.5points ahead
Where Qwen3.5 397B A17B pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+21points ahead
- Questions about very long textsLongBench v2+8.7points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+3.4points ahead
Every test, side by side
All 40 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.2
- SWE-bench ProQwen3.5 397B A17B by 2129.950.9+21
- SciCodeGPT-5.2 by 7.352.144.8+7.3
- Terminal-Bench HardGPT-5.2 by 6.14740.9+6.1
- SWE-bench VerifiedGPT-5.2 by 3.68076.4+3.6
- SWE-bench MultilingualQwen3.5 397B A17B by 2.666.769.3+2.6
- LMArena · WebDevGPT-5.2 by 16 rating points14151399+16 rating
AgentsEven
- GDPValGPT-5.2 by 33.548.314.8+33.5
- AA ApexAgentsGPT-5.2 by 7.72315.3+7.7
- BrowseCompQwen3.5 397B A17B by 3.265.869+3.2
- τ-Bench V3 · BankingQwen3.5 397B A17B by 2.311.113.4+2.3
ReasoningGPT-5.2
- CritPtGPT-5.2 by 9.911.61.7+9.9
- Humanity's Last ExamGPT-5.2 by 8.737.729+8.7
- GPQA DiamondGPT-5.2 by 190.389.3+1
FactsGPT-5.2
- AA-Omniscience · AccuracyGPT-5.2 by 13.544.330.8+13.5
- SimpleQA VerifiedGPT-5.2 by 11.137.126+11.1
- AA-Omniscience · Non-hallucinationGPT-5.2 by 7.718.811.1+7.7
Images and chartsQwen3.5 397B A17B
- MMMU-ProQwen3.5 397B A17B by 2.774.677.3+2.7
- LMArena · VisionQwen3.5 397B A17B by 18 rating points12461264+18 rating
MathGPT-5.2
- FrontierMath Tiers 1-3 (v2)GPT-5.2 by 36.267.431.2+36.2
- HMMT Feb. 2026GPT-5.2 by 9.19787.9+9.1
- AIME 2026GPT-5.2 by 798.391.3+7
- IMOAnswerBenchGPT-5.2 by 5.486.380.9+5.4
Long documentsEven
- LongBench v2Qwen3.5 397B A17B by 8.754.563.2+8.7
- AA-LCRGPT-5.2 by 5.482.777.3+5.4
Following instructionsQwen3.5 397B A17B
- IFBenchQwen3.5 397B A17B by 3.475.478.8+3.4
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.
- AA Agentic IndexGPT-5.2 by 49.660.210.6+49.6
- AA-OmniscienceGPT-5.2 by 29.8-0.9-30.8+29.8
- AA IntelligenceGPT-5.2 by 1230.418.4+12
- τ²-Bench Telecom (AA run)Qwen3.5 397B A17B by 10.884.895.6+10.8
- BrowseComp (context management)Qwen3.5 397B A17B by 8.67078.6+8.6
- ToolathlonGPT-5.2 by 846.338.3+8
- HMMT 2025GPT-5.2 by 6.799.492.7+6.7
- GPQA (unspecified)GPT-5.2 by 5.392.487.1+5.3
- HMMT Feb. 2025GPT-5.2 by 5.210094.8+5.2
- HMMT Nov. 2025GPT-5.2 by 3.195.892.7+3.1
- HLE (with tools)Qwen3.5 397B A17B by 2.845.548.3+2.8
- Terminal-Bench 2.0GPT-5.2 by 1.55452.5+1.5
- MMLU-ProQwen3.5 397B A17B by 1.186.787.8+1.1
- MMMLUGPT-5.2 by 1.189.688.5+1.1
- Artificial Analysis Coding Indextie48.748.2tie
Questions people ask
Which is better, GPT-5.2 or Qwen3.5 397B A17B?
GPT-5.2 wins four of the eight areas we test: coding, reasoning, facts and math. Qwen3.5 397B A17B wins images and charts and following instructions, and costs 73% less. They are level on agents and long documents.
Which is better for coding?
GPT-5.2. It wins 4 of the 6 coding tests both models report; Qwen3.5 397B A17B wins 2.
Which is cheaper?
GPT-5.2 costs $1.75 per million input tokens and $14.00 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes Qwen3.5 397B A17B about 73% cheaper for the same work.
How do you compare the two?
We use the 40 benchmark tests both models have published scores on. The verdict counts the 25 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.