Model comparison
Gemini 2.5 Pro vs GPT-4
Gemini 2.5 Pro is the stronger model overall, scoring 45.0 to 29.1 on the Noometry Index.
Last verified . 28 shared benchmarks.
GoogleSummary
- They share 28 benchmarks with published results for both. Gemini 2.5 Pro scores higher in 8 categories and GPT-4 in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Gemini 2.5 Pro leads 56.0 to 18.4.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 84.7% for Gemini 2.5 Pro and 1.1% for GPT-4.
- Gemini 2.5 Pro is cheaper at $1.25 / $10 per million input/output tokens, against $30 / $60 for GPT-4.
- Gemini 2.5 Pro accepts more context: 1.05M tokens versus 8K.
Side by side
| Gemini 2.5 Pro | GPT-4 | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 45.0 | 29.1 |
| Released | 2025-03-25 | 2023-03-14 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 8K |
| Max output | 66K | 8K |
| Input $ / M tokens | $1.25 | $30 |
| Output $ / M tokens | $10 | $60 |
| Results tracked | 78 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 2.5 Pro leads
Gemini 2.5 Pro: 42.4 (#101), GPT-4: 31.6 (#283)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| WeirdML | 54% | 12.4% |
| LMArena Coding | 1452 | 1254 |
| SWE-bench Verified | 57.6% | — |
| SWE-bench Verified (bash only) | 53.6% | — |
| Aider Polyglot | 83.1% | — |
| LMArena WebDev | 1227 | — |
| SciCode | 42.8% | — |
| GSO | 3.9% | — |
| BigCodeBench Instruct | — | 46% |
| LiveBench Coding | 85.9% | — |
| BigCodeBench Complete | — | 57.2% |
| CadEval | 64% | — |
| ALE-Bench | 785.52 | — |
| AlgoTune | 1.51 | — |
| HumanEval+ | — | 79.3% |
Agentic & Tool Use Not comparable
Gemini 2.5 Pro: 29.2 (#88), GPT-4: —
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| METR Time Horizons | 55.4% | 36.1% |
| Terminal-Bench | 32.6% | — |
| GDPval | 23.3% | — |
| Remote Labor Index | 0.8% | — |
| TheAgentCompany | 30.3% | — |
| τ²-bench Banking | 13.7% | — |
| DeepResearch Bench | 42.8% | — |
| BALROG | 43.3% | — |
| LMArena Search | 1142 | — |
| Vending-Bench 2 | 573.64 | — |
Reasoning Gemini 2.5 Pro leads
Gemini 2.5 Pro: 28.8 (#99), GPT-4: 17.8 (#289)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| Chess Puzzles | 20% | 4% |
| LMArena Hard Prompts | 1455 | 1241 |
| DTBench | 82.4% | 62.7% |
| LMCA | 34.8% | 17.1% |
| Epoch Capabilities Index | 145.32 | 125.89 |
| ForecastBench | 61.3 | 57.8 |
| ARC-AGI-2 | 4.9% | — |
| SimpleBench | 62.4% | — |
| Kagi LLM Benchmark | 70.3% | — |
| ARC-AGI-1 | 41% | — |
| CritPt | 2% | — |
| EnigmaEval | 5.6% | — |
| LiveBench Reasoning | 89.8% | — |
| Mystery Game Puzzles | — | 12% |
| LiveBench Data Analysis | 79.9% | — |
| BIG-Bench Hard | — | 75.1% |
| HellaSwag | — | 95.3% |
| LiveBench | 82.3% | — |
| WinoGrande | — | 87.5% |
Math Gemini 2.5 Pro leads
Gemini 2.5 Pro: 32.5 (#213), GPT-4: 10.8 (#309)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 84.7% | 1.1% |
| LMArena Math | 1450 | 1269 |
| MATH Level 5 | 95.9% | 23% |
| FrontierMath (Tiers 1-3) | 24.6% | — |
| FrontierMath Tier 4 | 0% | — |
| Omni-MATH | 41.6% | — |
| LiveBench Math | 90.2% | — |
| FrontierMath (Feb 2025 set) | 14.1% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |
| GSM8K | — | 92% |
Knowledge Gemini 2.5 Pro leads
Gemini 2.5 Pro: 56.0 (#46), GPT-4: 18.4 (#282)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| GPQA Diamond | 85.3% | 35.7% |
| LMArena Expert | 1452 | 1211 |
| Humanity's Last Exam | 21.6% | — |
| MMLU-Pro | 86.3% | — |
| Confabulations | 10.6% | — |
| Vectara Hallucination Rate | 7% | — |
| GPQA (HELM) | 74.9% | — |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |
Multimodal Not comparable
Gemini 2.5 Pro: 45.2 (#18), GPT-4: —
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| LMArena Vision | 1263 | — |
| GeoBench | 86% | — |
| VPCT | 48% | — |
| LMArena Document | 1421 | — |
| SpatialViz-Bench | 44.7% | — |
Multilingual Gemini 2.5 Pro leads
Gemini 2.5 Pro: 55.3 (#31), GPT-4: 40.6 (#215)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| LMArena Non-English | 1451 | 1246 |
| LMArena Chinese | 1507 | 1242 |
| LMArena French | 1472 | 1283 |
| LMArena German | 1487 | 1251 |
| LMArena Japanese | 1461 | 1209 |
| LMArena Korean | 1434 | 1184 |
| LMArena Russian | 1461 | 1251 |
| LMArena Spanish | 1473 | 1261 |
Instruction Following Gemini 2.5 Pro leads
Gemini 2.5 Pro: 75.0 (#75), GPT-4: 65.3 (#222)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1437 | 1241 |
| LiveBench Instruction Following | 80.6% | — |
| IFEval | 84% | — |
Long Context Gemini 2.5 Pro leads
Gemini 2.5 Pro: 59.8 (#5), GPT-4: 37.7 (#212)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1449 | 1244 |
| Fiction.LiveBench | 91.7% | — |
Writing & Preference Gemini 2.5 Pro leads
Gemini 2.5 Pro: 63.7 (#62), GPT-4: 34.9 (#268)
| Benchmark | Gemini 2.5 Pro | GPT-4 |
|---|---|---|
| LMArena Text | 1458 | 1263 |
| LMArena Creative Writing | 1454 | 1244 |
| EQ-Bench Creative Writing | 1421 | 752 |
| LMArena Multi-Turn | 1453 | 1257 |
| Short-Story Creative Writing | 83.8% | — |
| WildBench | 85.7% | — |
| LiveBench Language | 67.8% | — |
Frequently asked questions
Is Gemini 2.5 Pro better than GPT-4?
Gemini 2.5 Pro is the stronger model overall, scoring 45.0 to 29.1 on the Noometry Index.
Which is cheaper, Gemini 2.5 Pro or GPT-4?
Gemini 2.5 Pro is cheaper. It lists at $1.25 per million input tokens and $10 per million output tokens; GPT-4 lists at $30 and $60.
Is Gemini 2.5 Pro or GPT-4 better for coding?
Gemini 2.5 Pro scores higher on coding benchmarks: 42.4 versus 31.6 in the Noometry coding category.
Which has the bigger context window?
Gemini 2.5 Pro does, with 1.05M tokens against 8K.
How many benchmarks do Gemini 2.5 Pro and GPT-4 share?
28 benchmarks have published results for both models. Gemini 2.5 Pro has 78 scored results on Noometry and GPT-4 has 38.