GPT-5.2 vs Qwen3.6 Plus
Wins 3 of 8 areas
Reasoning · Math · Following instructions
Wins 2 of 8 areas
Agents · Facts
GPT-5.2 wins more areas, narrowly.Qwen3.6 Plus is cheaper and better at agents.
Scores updated · 51 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 math50GPT-5.25 of 5 tests
- ReasoningHard problems that need careful thinking40GPT-5.24 of 4 tests
- Following instructionsDoing exactly what it is asked10GPT-5.21 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own12Qwen3.6 Plus2 of 3 tests
- FactsGetting facts right instead of making them up12Qwen3.6 Plus2 of 3 tests
- CodingWriting and fixing software44Even4 each
- Images and chartsUnderstanding pictures, charts and video11Even1 each · 1 tie
- Long documentsFinding answers in very long texts11Even1 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.
Qwen3.6 Plus costs 78% 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- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+29.1points ahead
- Real work tasks from 44 professionsGDPVal+23.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+17.9points ahead
Where Qwen3.6 Plus pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+46.6points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+26.7points ahead
- Multi-step tasks using many real software toolsMCP Atlas+13.5points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SWE-bench ProQwen3.6 Plus by 26.729.956.6+26.7
- SciCodeGPT-5.2 by 11.452.140.7+11.4
- LiveBench · Agentic CodingGPT-5.2 by 8.950.341.4+8.9
- SWE-bench MultilingualQwen3.6 Plus by 7.166.773.8+7.1
- LMArena · WebDevQwen3.6 Plus by 45 rating points14161461+45 rating
- Terminal-Bench HardGPT-5.2 by 3.14743.9+3.1
- LiveBench · CodingQwen3.6 Plus by 2.176.178.2+2.1
- SWE-bench VerifiedGPT-5.2 by 1.28078.8+1.2
AgentsQwen3.6 Plus
- GDPValGPT-5.2 by 23.648.324.7+23.6
- MCP AtlasQwen3.6 Plus by 13.560.674.1+13.5
- τ-Bench V3 · BankingQwen3.6 Plus by 9.711.120.8+9.7
ReasoningGPT-5.2
- Humanity's Last ExamGPT-5.2 by 9.937.727.8+9.9
- CritPtGPT-5.2 by 8.711.62.9+8.7
- LiveBench · ReasoningGPT-5.2 by 7.483.275.8+7.4
- GPQA DiamondGPT-5.2 by 2.190.388.2+2.1
FactsQwen3.6 Plus
- AA-Omniscience · Non-hallucinationQwen3.6 Plus by 46.618.865.4+46.6
- AA-Omniscience · AccuracyGPT-5.2 by 17.944.326.4+17.9
- SimpleQA VerifiedQwen3.6 Plus by 737.144.1+7
Images and chartsEven
- MMMU-ProQwen3.6 Plus by 3.474.678+3.4
- Video-MMEGPT-5.2 by 1.88684.2+1.8
- CharXiv (RQ)tie82.181.5tie
MathGPT-5.2
- FrontierMath Tiers 1-3 (v2)GPT-5.2 by 29.167.438.3+29.1
- LiveBench · MathematicsGPT-5.2 by 9.593.283.7+9.5
- HMMT Feb. 2026GPT-5.2 by 9.29787.8+9.2
- AIME 2026GPT-5.2 by 398.395.3+3
- IMOAnswerBenchGPT-5.2 by 2.586.383.8+2.5
Long documentsEven
- LongBench v2Qwen3.6 Plus by 7.554.562+7.5
- AA-LCRGPT-5.2 by 4.482.778.3+4.4
Following instructionsGPT-5.2
- LiveBench · Instruction FollowingGPT-5.2 by 3.561.858.3+3.5
- IFBenchtie75.475.2tie
Other results21 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 31.260.229+31.2
- ScreenSpot-Pro (No tools)GPT-5.2 by 18.186.368.2+18.1
- τ²-Bench Telecom (AA run)Qwen3.6 Plus by 12.984.897.7+12.9
- SimpleVQAQwen3.6 Plus by 11.555.867.3+11.5
- frontiermath_tier_4_v1GPT-5.2 by 10.518.88.3+10.5
- livebench_data_analysisGPT-5.2 by 8.378.269.9+8.3
- Terminal-Bench 2.0Qwen3.6 Plus by 7.65461.6+7.6
- ToolathlonGPT-5.2 by 6.546.339.8+6.5
- Artificial Analysis Coding IndexQwen3.6 Plus by 5.848.754.5+5.8
- OmniDocBench 1.5Qwen3.6 Plus by 5.585.791.2+5.5
- HLE (with tools)Qwen3.6 Plus by 5.145.550.6+5.1
- MATH-VisionQwen3.6 Plus by 58388+5
- HMMT 2025GPT-5.2 by 4.899.494.6+4.8
- livebench_languageGPT-5.2 by 4.879.875+4.8
- LiveBenchGPT-5.2 by 3.974.870.8+3.9
- AA IntelligenceGPT-5.2 by 3.430.427+3.4
- VideoMMMUGPT-5.2 by 1.985.984+1.9
- AA-OmniscienceQwen3.6 Plus by 1.8-0.90.9+1.8
- MMLU-ProQwen3.6 Plus by 1.886.788.5+1.8
- HMMT Nov. 2025GPT-5.2 by 1.295.894.6+1.2
- MMMLUtie89.689.5tie
Questions people ask
Which is better, GPT-5.2 or Qwen3.6 Plus?
GPT-5.2 wins three of the eight areas we test: reasoning, math and following instructions. Qwen3.6 Plus wins agents and facts, and costs 78% less. They are level on coding, images and charts and long documents.
Which is better for coding?
Neither. They win 4 coding tests each of the 8 both models report.
Which is cheaper?
GPT-5.2 costs $1.75 per million input tokens and $14.00 per million output tokens; Qwen3.6 Plus costs $0.50 and $3.00. That makes Qwen3.6 Plus about 78% cheaper for the same work.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 21 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.