GPT-5 vs GPT-5.2
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
GPT-5.2 is the stronger all-rounder.GPT-5 is cheaper.
Scores updated · 59 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.
- CodingWriting and fixing software14GPT-5.24 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking14GPT-5.24 of 5 tests
- MathCompetition and research-level math03GPT-5.23 of 3 tests
- FactsGetting facts right instead of making them up13GPT-5.23 of 4 tests
- Images and chartsUnderstanding pictures, charts and video12GPT-5.22 of 4 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own12GPT-5.22 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.21 of 1 test
- Following instructionsDoing exactly what it is asked01GPT-5.21 of 1 test
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.
GPT-5 costs 29% 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
- Abstract visual puzzles that people can solveARC-AGI-2+43points ahead
- Real work tasks from 44 professionsGDPVal+27.3points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+14.4points ahead
Where GPT-5 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+13points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+11.9points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+11points ahead
Every test, side by side
All 59 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.2
- Terminal-Bench HardGPT-5.2 by 14.432.647+14.4
- SWE-bench ProGPT-5 by 11.941.829.9+11.9
- SWE-bench MultilingualGPT-5.2 by 11.455.366.7+11.4
- SciCodeGPT-5.2 by 9.242.952.1+9.2
- SWE-bench VerifiedGPT-5.2 by 5.174.980+5.1
- LMArena · WebDevtie14181416tie
AgentsGPT-5.2
- GDPValGPT-5.2 by 27.32148.3+27.3
- τ-Bench V3 · BankingGPT-5 by 1122.111.1+11
- BrowseCompGPT-5.2 by 10.954.965.8+10.9
ReasoningGPT-5.2
- ARC-AGI-2GPT-5.2 by 439.952.9+43
- SimpleBenchGPT-5 by 10.956.745.8+10.9
- Humanity's Last ExamGPT-5.2 by 9.228.537.7+9.2
- CritPtGPT-5.2 by 5.95.711.6+5.9
- GPQA DiamondGPT-5.2 by 4.985.490.3+4.9
FactsGPT-5.2
- SimpleQA VerifiedGPT-5 by 1350.137.1+13
- Vectara HHEM hallucination ratelower is betterGPT-5.2 by 4.315.110.8+4.3
- AA-Omniscience · AccuracyGPT-5.2 by 440.344.3+4
- AA-Omniscience · Non-hallucinationGPT-5.2 by 117.818.8+1
Images and chartsGPT-5.2
- OCRBenchv2GPT-5 by 555.550.5+5
- LMArena · VisionGPT-5.2 by 38 rating points12081246+38 rating
- CharXiv (RQ)GPT-5.2 by 181.182.1+1
- MMMU-Protie74.274.6tie
MathGPT-5.2
- FrontierMath Tiers 1-3 (v2)GPT-5.2 by 1255.467.4+12
- IMOAnswerBenchGPT-5.2 by 10.37686.3+10.3
- FrontierMath Tier 4GPT-5.2 by 9.721.931.7+9.7
Other results32 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 33.726.560.2+33.7
- ARC-AGI-1GPT-5.2 by 20.565.786.2+20.5
- Terminal-Bench 2.0GPT-5.2 by 18.835.254+18.8
- FrontierMath (overall)GPT-5.2 by 1426.340.3+14
- HMMT Feb. 2025GPT-5.2 by 11.788.3100+11.7
- Artificial Analysis Coding IndexGPT-5.2 by 10.937.848.7+10.9
- HLE (with tools)GPT-5.2 by 10.335.245.5+10.3
- metr_hcastGPT-5.2 by 8.668.477+8.6
- AA-OmniscienceGPT-5.2 by 7.8-8.7-0.9+7.8
- matharena_visual_math_overallGPT-5.2 by 7.778.886.5+7.7
- AA IntelligenceGPT-5.2 by 7.42330.4+7.4
- GPQA (unspecified)GPT-5.2 by 6.785.792.4+6.7
- HMMT Nov. 2025GPT-5.2 by 6.689.295.8+6.6
- frontiermath_tier_4_v1GPT-5.2 by 6.312.518.8+6.3
- HMMT 2025GPT-5.2 by 6.193.399.4+6.1
- LiveCodeBenchGPT-5.2 by 5.583.589+5.5
- AIME 2025GPT-5.2 by 5.494.6100+5.4
- metr_re_benchGPT-5 by 54035+5
- vectara_factual_consistencyGPT-5.2 by 4.384.989.2+4.3
- swe_bench_bashGPT-5.2 by 46569+4
- AIME25 no toolsGPT-5.2 by 3.494.698+3.4
- τ²-BenchGPT-5.2 by 3.182.485.5+3.1
- vectara_avg_summary_lengthGPT-5.2 by 23.6 rating points162.7186.3+23.6 rating
- VideoMMMUGPT-5.2 by 1.384.685.9+1.3
- HealthBench Consensus length-adjustedGPT-5 by 1.295.694.4+1.2
- HealthBench length-adjustedtie57.756.8tie
- τ²-Bench (Retail)tie81.182tie
- HealthBench Hard length-adjustedtie34.734.3tie
- MMLU-Protie87.186.7tie
- HealthBench Professional length-adjustedtie46.245.9tie
- vectara_answer_ratetie99.9100tie
- τ²-Bench Telecom (AA run)tie84.884.8tie
Questions people ask
Which is better, GPT-5 or GPT-5.2?
GPT-5.2 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. GPT-5 wins none, but costs 29% less.
Which is better for coding?
GPT-5.2. It wins 4 of the 6 coding tests both models report; GPT-5 wins 1, and 1 is a tie.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; GPT-5.2 costs $1.75 and $14.00. That makes GPT-5 about 29% cheaper for the same work.
How do you compare the two?
We use the 59 benchmark tests both models have published scores on. The verdict counts the 27 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 32 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.