Model comparison
DeepSeek-V3.2-Exp vs Grok 4
Grok 4 is the stronger model overall, scoring 48.1 to 44.3 on the Noometry Index.
Last verified . 31 shared benchmarks.
DeepSeekSummary
- They share 31 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 3 categories and Grok 4 in 6 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where Grok 4 leads 63.1 to 47.6.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 52.2% for DeepSeek-V3.2-Exp and 73.6% for Grok 4.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | Grok 4 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 44.3 | 48.1 |
| Released | 2025-09-29 | 2025-07-09 |
| Weights | Open | Proprietary |
| Context window | 164K | — |
| Max output | 66K | — |
| Input $ / M tokens | $0.26 | — |
| Output $ / M tokens | $0.38 | — |
| Results tracked | 49 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4 leads
DeepSeek-V3.2-Exp: 46.5 (#65), Grok 4: 50.3 (#46)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| Aider Polyglot | 74.2% | 79.6% |
| WeirdML | 39.5% | 45.7% |
| LMArena Coding | 1454 | 1408 |
| SWE-bench Verified (bash only) | 70% | — |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| SciCode | 38.9% | — |
Agentic & Tool Use Too close to call
DeepSeek-V3.2-Exp: 32.7 (#59), Grok 4: 32.3 (#68)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| Terminal-Bench | 39.6% | 27.2% |
| Berkeley Function Calling Leaderboard | 56.7% | 63% |
| APEX-Agents | 21.3% | — |
| GDPval | — | 21.1% |
| TheAgentCompany | 42.9% | — |
| Cybench | — | 43% |
| DeepResearch Bench | — | 47.3% |
| BALROG | — | 43.6% |
| LMArena Search | — | 1142 |
| METR Time Horizons | — | 66.6% |
| Vending-Bench 2 | 1,034 | — |
Reasoning Grok 4 leads
DeepSeek-V3.2-Exp: 22.1 (#208), Grok 4: 36.7 (#65)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| ARC-AGI-2 | 4% | 16% |
| Kagi LLM Benchmark | 52.2% | 73.6% |
| ARC-AGI-1 | 57% | 66.7% |
| Chess Puzzles | 14% | 28% |
| LMArena Hard Prompts | 1434 | 1409 |
| Epoch Capabilities Index | 146.27 | 146.44 |
| SimpleBench | — | 60.5% |
| NYT Connections (extended) | 36.7% | — |
| CritPt | 2.9% | — |
| Thematic Generalization | 65% | — |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
| ForecastBench | — | 60.9 |
Math Grok 4 leads
DeepSeek-V3.2-Exp: 41.7 (#87), Grok 4: 48.4 (#64)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 84% |
| LMArena Math | 1435 | 1422 |
| FrontierMath (Feb 2025 set) | 22.1% | 19.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 2.1% |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| Omni-MATH | — | 60.3% |
Knowledge Grok 4 leads
DeepSeek-V3.2-Exp: 51.7 (#66), Grok 4: 53.8 (#55)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| GPQA Diamond | 83.4% | 87% |
| LMArena Expert | 1436 | 1415 |
| MMLU-Pro | — | 85.1% |
| Confabulations | — | 12.4% |
| Vectara Hallucination Rate | 5.3% | — |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Grok 4: 33.7 (#94)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| LMArena Vision | — | 1210 |
| GeoBench | — | 45% |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.2 (#90), Grok 4: 51.8 (#103)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| LMArena Non-English | 1409 | 1403 |
| LMArena Chinese | 1461 | 1427 |
| LMArena French | 1433 | 1418 |
| LMArena German | 1440 | 1429 |
| LMArena Japanese | 1374 | 1394 |
| LMArena Korean | 1371 | 1377 |
| LMArena Russian | 1424 | 1410 |
| LMArena Spanish | 1440 | 1420 |
Instruction Following Grok 4 leads
DeepSeek-V3.2-Exp: 74.5 (#93), Grok 4: 79.2 (#5)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1387 |
| IFEval | — | 94.9% |
Long Context Grok 4 leads
DeepSeek-V3.2-Exp: 47.6 (#16), Grok 4: 63.1 (#4)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| Fiction.LiveBench | 83.3% | 94.4% |
| LMArena Longer Query | 1428 | 1409 |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.4 (#77), Grok 4: 58.5 (#116)
| Benchmark | DeepSeek-V3.2-Exp | Grok 4 |
|---|---|---|
| LMArena Text | 1425 | 1411 |
| LMArena Creative Writing | 1403 | 1397 |
| LMArena Multi-Turn | 1427 | 1416 |
| Short-Story Creative Writing | — | 76.9% |
| EQ-Bench Creative Writing | 1515 | — |
| WildBench | — | 79.7% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Grok 4?
Grok 4 is the stronger model overall, scoring 48.1 to 44.3 on the Noometry Index.
Is DeepSeek-V3.2-Exp or Grok 4 better for coding?
Grok 4 scores higher on coding benchmarks: 50.3 versus 46.5 in the Noometry coding category.
How many benchmarks do DeepSeek-V3.2-Exp and Grok 4 share?
31 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Grok 4 has 48.