Model comparison
DeepSeek-V3 vs Grok 4
Grok 4 is the stronger model overall, scoring 48.1 to 39.5 on the Noometry Index.
Last verified . 35 shared benchmarks.
DeepSeekSummary
- They share 35 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and Grok 4 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where Grok 4 leads 63.1 to 34.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 84% for Grok 4.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | Grok 4 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 39.5 | 48.1 |
| Released | 2024-12-26 | 2025-07-09 |
| Weights | Open | Proprietary |
| Context window | 164K | — |
| Max output | 164K | — |
| Input $ / M tokens | $0.24 | — |
| Output $ / M tokens | $0.90 | — |
| Results tracked | 60 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4 leads
DeepSeek-V3: 42.3 (#106), Grok 4: 50.3 (#46)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| Aider Polyglot | 55.1% | 79.6% |
| WeirdML | 36.1% | 45.7% |
| LMArena Coding | 1368 | 1408 |
| SciCode | 35.8% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Grok 4: 32.3 (#68)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| METR Time Horizons | 49.6% | 66.6% |
| Terminal-Bench | — | 27.2% |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 21.1% |
| Cybench | — | 43% |
| DeepResearch Bench | — | 47.3% |
| BALROG | — | 43.6% |
| LMArena Search | — | 1142 |
Reasoning Grok 4 leads
DeepSeek-V3: 20.5 (#236), Grok 4: 36.7 (#65)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| SimpleBench | 27.2% | 60.5% |
| Kagi LLM Benchmark | 52.3% | 73.6% |
| LMArena Hard Prompts | 1365 | 1409 |
| Epoch Capabilities Index | 135.94 | 146.44 |
| ForecastBench | 59.1 | 60.9 |
| ARC-AGI-2 | — | 16% |
| ARC-AGI-1 | — | 66.7% |
| CritPt | 0% | — |
| Chess Puzzles | — | 28% |
| LiveBench Reasoning | 65.8% | — |
| DTBench | 64.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| LMCA | 15.5% | — |
| BIG-Bench Hard | 87.5% | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math Grok 4 leads
DeepSeek-V3: 32.1 (#219), Grok 4: 48.4 (#64)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 84% |
| Omni-MATH | 40.3% | 60.3% |
| LMArena Math | 1373 | 1422 |
| FrontierMath (Feb 2025 set) | 1.7% | 19.7% |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge Grok 4 leads
DeepSeek-V3: 37.5 (#155), Grok 4: 53.8 (#55)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| GPQA Diamond | 67.6% | 87% |
| MMLU-Pro | 72.3% | 85.1% |
| Confabulations | 26.1% | 12.4% |
| GPQA (HELM) | 53.8% | 72.7% |
| LMArena Expert | 1351 | 1415 |
| Vectara Hallucination Rate | 6.1% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, Grok 4: 33.7 (#94)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| LMArena Vision | — | 1210 |
| GeoBench | — | 45% |
Multilingual Grok 4 leads
DeepSeek-V3: 48.5 (#143), Grok 4: 51.8 (#103)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| LMArena Non-English | 1358 | 1403 |
| LMArena Chinese | 1391 | 1427 |
| LMArena French | 1385 | 1418 |
| LMArena German | 1374 | 1429 |
| LMArena Japanese | 1333 | 1394 |
| LMArena Korean | 1319 | 1377 |
| LMArena Russian | 1373 | 1410 |
| LMArena Spanish | 1358 | 1420 |
Instruction Following Grok 4 leads
DeepSeek-V3: 72.8 (#130), Grok 4: 79.2 (#5)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| IFEval | 83.2% | 94.9% |
| LMArena Instruction Following | 1345 | 1387 |
| LiveBench Instruction Following | 81.5% | — |
Long Context Grok 4 leads
DeepSeek-V3: 34.0 (#253), Grok 4: 63.1 (#4)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| Fiction.LiveBench | 50% | 94.4% |
| LMArena Longer Query | 1352 | 1409 |
Writing & Preference Grok 4 leads
DeepSeek-V3: 57.4 (#130), Grok 4: 58.5 (#116)
| Benchmark | DeepSeek-V3 | Grok 4 |
|---|---|---|
| LMArena Text | 1375 | 1411 |
| LMArena Creative Writing | 1364 | 1397 |
| Short-Story Creative Writing | 77% | 76.9% |
| WildBench | 83% | 79.7% |
| LMArena Multi-Turn | 1389 | 1416 |
| EQ-Bench Creative Writing | 1472 | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Grok 4?
Grok 4 is the stronger model overall, scoring 48.1 to 39.5 on the Noometry Index.
Is DeepSeek-V3 or Grok 4 better for coding?
Grok 4 scores higher on coding benchmarks: 50.3 versus 42.3 in the Noometry coding category.
How many benchmarks do DeepSeek-V3 and Grok 4 share?
35 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Grok 4 has 48.