Model comparison
DeepSeek V4 Pro vs Grok 4
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 48.1 on the Noometry Index.
Last verified . 26 shared benchmarks.
DeepSeekSummary
- They share 26 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 7 categories and Grok 4 in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 36.7.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 16% for Grok 4.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Pro | Grok 4 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 54.3 | 48.1 |
| Released | 2026-04-24 | 2025-07-09 |
| Weights | Open | Proprietary |
| Context window | 1M | — |
| Max output | 393K | — |
| Input $ / M tokens | $0.66 | — |
| Output $ / M tokens | $1.98 | — |
| Results tracked | 48 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), Grok 4: 50.3 (#46)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| WeirdML | 66.2% | 45.7% |
| LMArena Coding | 1470 | 1408 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| Aider Polyglot | — | 79.6% |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use Too close to call
DeepSeek V4 Pro: 32.8 (#58), Grok 4: 32.3 (#68)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| Terminal-Bench | — | 27.2% |
| APEX-Agents | 47.3% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 21.1% |
| Cybench | — | 43% |
| DeepResearch Bench | — | 47.3% |
| BALROG | — | 43.6% |
| LMArena Search | — | 1142 |
| METR Time Horizons | — | 66.6% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Grok 4: 36.7 (#65)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 16% |
| Kagi LLM Benchmark | 53.5% | 73.6% |
| ARC-AGI-1 | 90.5% | 66.7% |
| Chess Puzzles | 47% | 28% |
| LMArena Hard Prompts | 1461 | 1409 |
| Epoch Capabilities Index | 155.31 | 146.44 |
| ForecastBench | 56.1 | 60.9 |
| SimpleBench | — | 60.5% |
| NYT Connections (extended) | 91.3% | — |
| CritPt | 18% | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Grok 4: 48.4 (#64)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 84% |
| LMArena Math | 1455 | 1422 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| Omni-MATH | — | 60.3% |
| FrontierMath (Feb 2025 set) | — | 19.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Grok 4: 53.8 (#55)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| GPQA Diamond | 91.7% | 87% |
| LMArena Expert | 1464 | 1415 |
| SimpleQA Verified | 52.9% | — |
| MMLU-Pro | — | 85.1% |
| Confabulations | — | 12.4% |
| Vectara Hallucination Rate | 8.6% | — |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
DeepSeek V4 Pro: —, Grok 4: 33.7 (#94)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| LMArena Vision | — | 1210 |
| GeoBench | — | 45% |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), Grok 4: 51.8 (#103)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| LMArena Non-English | 1439 | 1403 |
| LMArena Chinese | 1486 | 1427 |
| LMArena French | 1472 | 1418 |
| LMArena German | 1458 | 1429 |
| LMArena Japanese | 1445 | 1394 |
| LMArena Korean | 1447 | 1377 |
| LMArena Russian | 1453 | 1410 |
| LMArena Spanish | 1458 | 1420 |
Instruction Following Grok 4 leads
DeepSeek V4 Pro: 76.1 (#47), Grok 4: 79.2 (#5)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1387 |
| IFEval | — | 94.9% |
Long Context Grok 4 leads
DeepSeek V4 Pro: 45.0 (#51), Grok 4: 63.1 (#4)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| LMArena Longer Query | 1458 | 1409 |
| Fiction.LiveBench | — | 94.4% |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), Grok 4: 58.5 (#116)
| Benchmark | DeepSeek V4 Pro | Grok 4 |
|---|---|---|
| LMArena Text | 1451 | 1411 |
| LMArena Creative Writing | 1446 | 1397 |
| LMArena Multi-Turn | 1467 | 1416 |
| Short-Story Creative Writing | — | 76.9% |
| EQ-Bench Creative Writing | 1553 | — |
| WildBench | — | 79.7% |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than Grok 4?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 48.1 on the Noometry Index.
Is DeepSeek V4 Pro or Grok 4 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 50.3 in the Noometry coding category.
How many benchmarks do DeepSeek V4 Pro and Grok 4 share?
26 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Grok 4 has 48.