Gemini 3 Pro vs GPT-5.2
Wins 1 of 8 areas
Images and charts
Wins 4 of 8 areas
Agents · Math · Long documents · Following instructions
GPT-5.2 wins more areas, narrowly.Gemini 3 Pro is cheaper and better at images and charts.
Scores updated · 78 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.
- AgentsCarrying out multi-step tasks on its own04GPT-5.24 of 4 tests
- MathCompetition and research-level math03GPT-5.23 of 3 tests
- Long documentsFinding answers in very long texts12GPT-5.22 of 3 tests
- Following instructionsDoing exactly what it is asked01GPT-5.21 of 1 test
- Images and chartsUnderstanding pictures, charts and video60Gemini 3 Pro6 of 7 tests · 1 tie
- CodingWriting and fixing software33Even3 each
- ReasoningHard problems that need careful thinking22Even2 each · 1 tie
- FactsGetting facts right instead of making them up22Even2 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.
Gemini 3 Pro costs 11% 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+21.8points ahead
- Real work tasks from 44 professionsGDPVal+14.1points ahead
- Harvard-MIT high-school math contest problemsHMMT Feb. 2026+10.6points ahead
Where Gemini 3 Pro pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+35.8points ahead
- Common-sense trick questionsSimpleBench+30.6points ahead
- Questions about very long textsLongBench v2+13.7points ahead
Every test, side by side
All 78 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SWE-bench ProGemini 3 Pro by 13.443.329.9+13.4
- Terminal-Bench HardGPT-5.2 by 5.341.747+5.3
- SciCodeGemini 3 Pro by 456.152.1+4
- SWE-bench VerifiedGPT-5.2 by 3.876.280+3.8
- LMArena · WebDevGemini 3 Pro by 24 rating points14391415+24 rating
- SWE-bench MultilingualGPT-5.2 by 1.76566.7+1.7
AgentsGPT-5.2
- GDPValGPT-5.2 by 14.134.248.3+14.1
- BrowseCompGPT-5.2 by 6.659.265.8+6.6
- MCP AtlasGPT-5.2 by 6.554.160.6+6.5
- AA ApexAgentsGPT-5.2 by 4.618.423+4.6
ReasoningEven
- SimpleBenchGemini 3 Pro by 30.676.445.8+30.6
- ARC-AGI-2GPT-5.2 by 21.831.152.9+21.8
- CritPtGPT-5.2 by 2.59.111.6+2.5
- Humanity's Last ExamGemini 3 Pro by 239.737.7+2
- GPQA Diamondtie90.890.3tie
FactsEven
- SimpleQA VerifiedGemini 3 Pro by 35.872.937.1+35.8
- AA-Omniscience · AccuracyGemini 3 Pro by 11.555.844.3+11.5
- AA-Omniscience · Non-hallucinationGPT-5.2 by 10.38.518.8+10.3
- Vectara HHEM hallucination ratelower is betterGPT-5.2 by 2.813.610.8+2.8
Images and chartsGemini 3 Pro
- OCRBenchv2Gemini 3 Pro by 12.963.450.5+12.9
- OCRBenchGemini 3 Pro by 9.690.380.7+9.6
- MathVistaGemini 3 Pro by 789.882.8+7
- LMArena · VisionGemini 3 Pro by 59 rating points13051246+59 rating
- MMMU-ProGemini 3 Pro by 5.680.274.6+5.6
- Video-MMEGemini 3 Pro by 2.488.486+2.4
- CharXiv (RQ)tie81.482.1tie
MathGPT-5.2
- HMMT Feb. 2026GPT-5.2 by 10.686.497+10.6
- AIME 2026GPT-5.2 by 6.691.798.3+6.6
- IMOAnswerBenchGPT-5.2 by 3.283.186.3+3.2
Long documentsGPT-5.2
- LongBench v2Gemini 3 Pro by 13.768.254.5+13.7
- MRCR v2 8 needle 128k (average)GPT-5.2 by 6.87783.8+6.8
- AA-LCRGPT-5.2 by 6.77682.7+6.7
Other results45 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.
- InfoVQA (val)GPT-5.2 by 26.857.284+26.8
- GDPval-AA (Elo)GPT-5.2 by 267 rating points11951462+267 rating
- WorldVQAGemini 3 Pro by 19.447.428+19.4
- AA-OmniscienceGemini 3 Pro by 16.215.3-0.9+16.2
- SimpleVQAGemini 3 Pro by 13.969.755.8+13.9
- ScreenSpot-Pro (No tools)GPT-5.2 by 13.672.786.3+13.6
- τ²-BenchGemini 3 Pro by 12.59885.5+12.5
- ARC-AGI-1GPT-5.2 by 11.27586.2+11.2
- BrowseComp (context management)GPT-5.2 by 10.859.270+10.8
- ToolathlonGPT-5.2 by 9.936.446.3+9.9
- vectara_avg_summary_lengthGPT-5.2 by 84.4 rating points101.9186.3+84.4 rating
- AA Agentic IndexGPT-5.2 by 8.25260.2+8.2
- DeepSearchQA (F1)GPT-5.2 by 8.163.271.3+8.1
- metr_hcastGPT-5.2 by 5.771.377+5.7
- MotionBenchGemini 3 Pro by 5.570.364.8+5.5
- ZeroBench (w/ tools)Gemini 3 Pro by 5127+5
- LiveCodeBench Pro (Elo)Gemini 3 Pro by 46 rating points24392393+46 rating
- MMLU-ProGemini 3 Pro by 3.490.186.7+3.4
- MMVUGPT-5.2 by 3.377.580.8+3.3
- τ²-Bench (Retail)Gemini 3 Pro by 3.385.382+3.3
- MATH-VisionGemini 3 Pro by 3.186.183+3.1
- LiveCodeBenchGemini 3 Pro by 39289+3
- SWE-PerfGemini 3 Pro by 2.96.53.6+2.9
- OmniDocBench 1.5Gemini 3 Pro by 2.888.585.7+2.8
- vectara_factual_consistencyGPT-5.2 by 2.886.489.2+2.8
- HMMT Feb. 2025GPT-5.2 by 2.597.5100+2.5
- HMMT Nov. 2025GPT-5.2 by 2.593.395.8+2.5
- AA IntelligenceGPT-5.2 by 2.42830.4+2.4
- matharena_visual_math_overallGPT-5.2 by 2.384.286.5+2.3
- τ²-Bench Telecom (AA run)Gemini 3 Pro by 2.387.184.8+2.3
- Artificial Analysis Coding IndexGPT-5.2 by 2.246.548.7+2.2
- MMMLUGemini 3 Pro by 2.291.889.6+2.2
- AIME25 no toolsGPT-5.2 by 29698+2
- VideoMMMUGemini 3 Pro by 1.787.685.9+1.7
- LiveBenchGPT-5.2 by 1.473.474.8+1.4
- LongVideoBenchGemini 3 Pro by 1.277.776.5+1.2
- SWT-benchGPT-5.2 by 179.780.7+1
- ZeroBenchGPT-5.2 by 189+1
- vectara_answer_ratetie99.4100tie
- HLE (with tools)tie45.845.5tie
- Terminal-Bench 2.0tie54.254tie
- metr_swaatie99.599.6tie
- AIME 2025tie100100tie
- frontiermath_tier_4_v1tie18.818.8tie
- metr_re_benchtie3535tie
Questions people ask
Which is better, Gemini 3 Pro or GPT-5.2?
GPT-5.2 wins four of the eight areas we test: agents, math, long documents and following instructions. Gemini 3 Pro wins images and charts, and costs 11% less. They are level on coding, reasoning and facts.
Which is better for coding?
Neither. They win 3 coding tests each of the 6 both models report.
Which is cheaper?
Gemini 3 Pro costs $2.00 per million input tokens and $12.00 per million output tokens; GPT-5.2 costs $1.75 and $14.00. That makes Gemini 3 Pro about 11% cheaper for the same work.
How do you compare the two?
We use the 78 benchmark tests both models have published scores on. The verdict counts the 33 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 45 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.