Model comparison
Grok 4 vs o3
Grok 4 and o3 score almost the same on the Noometry Index (48.1 vs 47.5), so choose on price, context window or the category you care about most.
Last verified . 45 shared benchmarks.
Summary
- They share 45 benchmarks with published results for both. Grok 4 scores higher in 5 categories and o3 in 5 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where Grok 4 leads 63.1 to 53.3.
- The biggest single-benchmark swing is GeoBench: 45% for Grok 4 and 74% for o3.
Side by side
| Grok 4 | o3 | |
|---|---|---|
| Provider | xAI | OpenAI |
| Noometry Index | 48.1 | 47.5 |
| Released | 2025-07-09 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | — | 200K |
| Max output | — | 100K |
| Input $ / M tokens | — | $2 |
| Output $ / M tokens | — | $8 |
| Results tracked | 48 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4 leads
Grok 4: 50.3 (#46), o3: 46.8 (#64)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| Aider Polyglot | 79.6% | 81.3% |
| WeirdML | 45.7% | 52.4% |
| LMArena Coding | 1408 | 1408 |
| SWE-bench Verified | — | 62.3% |
| SWE-bench Verified (bash only) | — | 58.4% |
| GSO | — | 8.8% |
| CadEval | — | 74% |
| ALE-Bench | — | 933.55 |
Agentic & Tool Use o3 leads
Grok 4: 32.3 (#68), o3: 34.5 (#44)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 63% | 63% |
| GDPval | 21.1% | 30.8% |
| DeepResearch Bench | 47.3% | 45.2% |
| LMArena Search | 1142 | 1144 |
| METR Time Horizons | 66.6% | 65.4% |
| Terminal-Bench | 27.2% | — |
| Cybench | 43% | — |
| OSWorld | — | 23% |
| BALROG | 43.6% | — |
Reasoning Grok 4 leads
Grok 4: 36.7 (#65), o3: 32.0 (#78)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| ARC-AGI-2 | 16% | 6.5% |
| SimpleBench | 60.5% | 53.1% |
| Kagi LLM Benchmark | 73.6% | 67.6% |
| ARC-AGI-1 | 66.7% | 60.8% |
| Chess Puzzles | 28% | 38% |
| LMArena Hard Prompts | 1409 | 1402 |
| Epoch Capabilities Index | 146.44 | 146.86 |
| ForecastBench | 60.9 | 62.5 |
| CritPt | — | 1.4% |
| EnigmaEval | — | 13.1% |
| Mystery Game Puzzles | — | 29% |
| DTBench | — | 84.8% |
| LMCA | — | 39.7% |
Math o3 leads
Grok 4: 48.4 (#64), o3: 50.2 (#58)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 84% | 84.4% |
| Omni-MATH | 60.3% | 71.4% |
| LMArena Math | 1422 | 1426 |
| FrontierMath (Feb 2025 set) | 19.7% | 18.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 2.1% |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| MATH Level 5 | — | 97.8% |
Knowledge Too close to call
Grok 4: 53.8 (#55), o3: 54.6 (#52)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| GPQA Diamond | 87% | 81.8% |
| MMLU-Pro | 85.1% | 85.9% |
| Confabulations | 12.4% | 14.4% |
| GPQA (HELM) | 72.7% | 75.3% |
| LMArena Expert | 1415 | 1402 |
| Humanity's Last Exam | — | 20.3% |
| SimpleQA Verified | — | 49.4% |
Multimodal o3 leads
Grok 4: 33.7 (#94), o3: 41.4 (#36)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| LMArena Vision | 1210 | 1214 |
| GeoBench | 45% | 74% |
| VPCT | — | 52% |
Multilingual Too close to call
Grok 4: 51.8 (#103), o3: 51.7 (#105)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| LMArena Non-English | 1403 | 1401 |
| LMArena Chinese | 1427 | 1437 |
| LMArena French | 1418 | 1430 |
| LMArena German | 1429 | 1420 |
| LMArena Japanese | 1394 | 1403 |
| LMArena Korean | 1377 | 1370 |
| LMArena Russian | 1410 | 1406 |
| LMArena Spanish | 1420 | 1395 |
Instruction Following Grok 4 leads
Grok 4: 79.2 (#5), o3: 72.8 (#127)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| IFEval | 94.9% | 86.9% |
| LMArena Instruction Following | 1387 | 1368 |
Long Context Grok 4 leads
Grok 4: 63.1 (#4), o3: 53.3 (#6)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| Fiction.LiveBench | 94.4% | 88.9% |
| LMArena Longer Query | 1409 | 1372 |
| CL-bench | — | 17.8% |
Writing & Preference o3 leads
Grok 4: 58.5 (#116), o3: 63.5 (#64)
| Benchmark | Grok 4 | o3 |
|---|---|---|
| LMArena Text | 1411 | 1410 |
| LMArena Creative Writing | 1397 | 1359 |
| Short-Story Creative Writing | 76.9% | 83.9% |
| WildBench | 79.7% | 86.1% |
| LMArena Multi-Turn | 1416 | 1405 |
| EQ-Bench Creative Writing | — | 1676 |
Frequently asked questions
Is Grok 4 better than o3?
Grok 4 and o3 score almost the same on the Noometry Index (48.1 vs 47.5), so choose on price, context window or the category you care about most.
Is Grok 4 or o3 better for coding?
Grok 4 scores higher on coding benchmarks: 50.3 versus 46.8 in the Noometry coding category.
How many benchmarks do Grok 4 and o3 share?
45 benchmarks have published results for both models. Grok 4 has 48 scored results on Noometry and o3 has 63.