Model comparison
Grok-2 (Dec 2024) vs o3-pro
o3-pro is the stronger model overall, scoring 42.9 to 33.7 on the Noometry Index.
Last verified . 5 shared benchmarks.
Summary
- They share 5 benchmarks with published results for both. Grok-2 (Dec 2024) scores higher in 1 category and o3-pro in 4 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in long context, where o3-pro leads 72.2 to 38.8.
- The biggest single-benchmark swing is WeirdML: 22.2% for Grok-2 (Dec 2024) and 58.2% for o3-pro.
Side by side
| Grok-2 (Dec 2024) | o3-pro | |
|---|---|---|
| Provider | xAI | OpenAI |
| Noometry Index | 33.7 | 42.9 |
| Released | 2024-08-13 | 2025-06-10 |
| Weights | Proprietary | Proprietary |
| Context window | — | 200K |
| Max output | — | 100K |
| Input $ / M tokens | — | $20 |
| Output $ / M tokens | — | $80 |
| Results tracked | 34 | 12 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3-pro leads
Grok-2 (Dec 2024): 33.3 (#258), o3-pro: 55.5 (#24)
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| WeirdML | 22.2% | 58.2% |
| Aider Polyglot | — | 84.9% |
| LiveBench Coding | 46.4% | — |
| LMArena Coding | 1287 | — |
Reasoning o3-pro leads
Grok-2 (Dec 2024): 16.9 (#299), o3-pro: 23.8 (#171)
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| DTBench | 65.2% | 86.9% |
| Epoch Capabilities Index | 130.48 | 147.42 |
| ARC-AGI-2 | — | 4.9% |
| SimpleBench | 22.7% | — |
| Kagi LLM Benchmark | — | 72.1% |
| ARC-AGI-1 | — | 59.3% |
| LiveBench Reasoning | 54.8% | — |
| LMArena Hard Prompts | 1272 | — |
| LiveBench Data Analysis | 54.5% | — |
| LMCA | — | 38.5% |
| LiveBench | 54.3% | — |
Math Not comparable
Grok-2 (Dec 2024): 20.8 (#284), o3-pro: —
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 11.5% | — |
| LiveBench Math | 54.9% | — |
| LMArena Math | 1283 | — |
| MATH Level 5 | 63.5% | — |
| FrontierMath (Feb 2025 set) | 0.7% | — |
Knowledge Too close to call
Grok-2 (Dec 2024): 29.8 (#233), o3-pro: 29.5 (#238)
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| Confabulations | 20.1% | 14.2% |
| GPQA Diamond | 53.8% | — |
| Vectara Hallucination Rate | — | 23.3% |
| LMArena Expert | 1254 | — |
Multilingual Not comparable
Grok-2 (Dec 2024): 43.1 (#188), o3-pro: —
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| LMArena Non-English | 1282 | — |
| LMArena Chinese | 1289 | — |
| LMArena French | 1318 | — |
| LMArena German | 1287 | — |
| LMArena Japanese | 1244 | — |
| LMArena Korean | 1237 | — |
| LMArena Russian | 1286 | — |
| LMArena Spanish | 1281 | — |
Instruction Following Not comparable
Grok-2 (Dec 2024): 66.9 (#202), o3-pro: —
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| LiveBench Instruction Following | 69.6% | — |
| LMArena Instruction Following | 1270 | — |
Long Context o3-pro leads
Grok-2 (Dec 2024): 38.8 (#190), o3-pro: 72.2 (#1)
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| Fiction.LiveBench | — | 97.2% |
| LMArena Longer Query | 1276 | — |
Writing & Preference o3-pro leads
Grok-2 (Dec 2024): 48.6 (#198), o3-pro: 57.1 (#133)
| Benchmark | Grok-2 (Dec 2024) | o3-pro |
|---|---|---|
| Short-Story Creative Writing | 63.6% | 84.4% |
| LMArena Text | 1305 | — |
| LMArena Creative Writing | 1284 | — |
| LMArena Multi-Turn | 1290 | — |
| LiveBench Language | 45.6% | — |
Frequently asked questions
Is Grok-2 (Dec 2024) better than o3-pro?
o3-pro is the stronger model overall, scoring 42.9 to 33.7 on the Noometry Index.
Is Grok-2 (Dec 2024) or o3-pro better for coding?
o3-pro scores higher on coding benchmarks: 55.5 versus 33.3 in the Noometry coding category.
How many benchmarks do Grok-2 (Dec 2024) and o3-pro share?
5 benchmarks have published results for both models. Grok-2 (Dec 2024) has 34 scored results on Noometry and o3-pro has 12.