The two are evenly matched.Gemini 1.5 Pro is better at reasoning; GPT-4o at images and charts.
Scores updated · 55 tests both models report · How we compare
Where each one wins
Tests won in each of the three 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.
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.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemini 1.5 Pro pulls ahead
- Math problems shown in pictures and chartsMathVista+6.7points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+6.3points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+2.8points ahead
Every test, side by side
All 55 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SciCodeGPT-4o by 3.829.533.3+3.8
- SWE-bench VerifiedGemini 1.5 Pro by 134.233.2+1
ReasoningGemini 1.5 Pro
- GPQA DiamondGemini 1.5 Pro by 6.358.952.6+6.3
- Humanity's Last ExamGemini 1.5 Pro by 2.84.61.8+2.8
Images and chartsGPT-4o
Other results47 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.
- M-LongDoc (multimodal long-document benchmark)GPT-4o by 15.226.241.4+15.2
- MMMU (val) (Pass@1)GPT-4o by 1554.169.1+15
- SimpleQAGPT-4o by 14.823.438.2+14.8
- DROPGPT-4o by 8.574.983.4+8.5
- DROP F1 ScoreGPT-4o by 8.574.983.4+8.5
- IFEvalGemini 1.5 Pro by 8.489.481+8.4
- LiveCodeBenchGPT-4o by 7.830.538.3+7.8
- Ruler 16kGemini 1.5 Pro by 79689+7
- Ruler 32kGemini 1.5 Pro by 795.888.8+7
- MMMU (validation)GPT-4o by 6.962.269.1+6.9
- MTOB eng → kalam (ChrF) no contextGemini 1.5 Pro by 6.916.89.9+6.9
- OlympiadBenchGemini 1.5 Pro by 6.932.125.2+6.9
- HumanEvalGPT-4o by 6.184.190.2+6.1
- HumanEval 0-shotGPT-4o by 6.184.190.2+6.1
- Arena HardGemini 1.5 Pro by 685.379.3+6
- MMLUGemini 1.5 Pro by 5.779.573.8+5.7
- RULER 64KGemini 1.5 Pro by 5.493.888.4+5.4
- IFEval (avg)Gemini 1.5 Pro by 5.389.484.1+5.3
- AIR-Bench 2024Gemini 1.5 Pro by 4.967.362.4+4.9
- naturalquestions_closedbookGPT-4o by 445.549.6+4
- NarrativeQAGPT-4o by 3.975.679.5+3.9
- Ruler 8kGemini 1.5 Pro by 3.99692.1+3.9
- MEGA-Bench_macroGPT-4o by 3.545.949.4+3.5
- MTOB kalam → eng (BLEURT) half bookGemini 1.5 Pro by 3.261.558.3+3.2
- Video-MME (wo subs)Gemini 1.5 Pro by 3.17571.9+3.1
- ChartQAGemini 1.5 Pro by 388.785.7+3
- HarmBenchGPT-4o by 379.982.9+3
- MGSMGPT-4o by 387.590.5+3
- MATHGPT-4o by 2.882.585.3+2.8
- OpenBookQAGPT-4o by 1.695.296.8+1.6
- XSTestGemini 1.5 Pro by 1.598.897.3+1.5
- DocVQAGPT-4o by 1.391.592.8+1.3
- MTOB kalam → eng (BLEURT) no contextGPT-4o by 1.23233.2+1.2
- MMLU-ProGemini 1.5 Pro by 1.175.874.7+1.1
- simple_safety_testsGPT-4o by 197.598.5+1
- anthropic_red_teamtie99.999.1tie
- MBPP+ (EvalPlus-augmented)tie75.476.2tie
- Ruler 4ktie96.297tie
- Chinese SimpleQA (C-SimpleQA)tie59.458.7tie
- AA Intelligencetie7.97.3tie
- Artificial Analysis Coding Indextie23.624.2tie
- ChartQA_relaxedtie88.788.1tie
- MTOB eng → kalam (ChrF) half booktie53.754.3tie
- DocVQA (test, ANLS score)tie93.192.8tie
- AI2Dtie94.494.2tie
- GSM8Ktie90.890.9tie
- DROP (F1)tie89.289.2tie
Questions people ask
Which is better, Gemini 1.5 Pro or GPT-4o?
Gemini 1.5 Pro and GPT-4o each win one of the three areas where both have results. Gemini 1.5 Pro wins reasoning; GPT-4o wins images and charts. They are level on coding.
Which is better for coding?
Neither. They win 1 coding test each of the 2 both models report.
How do you compare the two?
We use the 55 benchmark tests both models have published scores on. The verdict counts the 8 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 47 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.