Model comparison
DeepSeek-V2.5 (Sep 2024) vs Grok 4
Grok 4 is the stronger model overall, scoring 48.1 to 37.6 on the Noometry Index.
Last verified . 18 shared benchmarks.
DeepSeekSummary
- They share 18 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) 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 39.5.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 79.6% for Grok 4.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Grok 4 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 37.6 | 48.1 |
| Released | 2024-09-06 | 2025-07-09 |
| Weights | Open | Proprietary |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Grok 4: 50.3 (#46)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| Aider Polyglot | 17.8% | 79.6% |
| LMArena Coding | 1309 | 1408 |
| WeirdML | — | 45.7% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Grok 4: 32.3 (#68)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| Terminal-Bench | — | 27.2% |
| 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% |
Reasoning Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Grok 4: 36.7 (#65)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1409 |
| ARC-AGI-2 | — | 16% |
| SimpleBench | — | 60.5% |
| Kagi LLM Benchmark | — | 73.6% |
| ARC-AGI-1 | — | 66.7% |
| Chess Puzzles | — | 28% |
| Epoch Capabilities Index | — | 146.44 |
| ForecastBench | — | 60.9 |
Math Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Grok 4: 48.4 (#64)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Math | 1288 | 1422 |
| OTIS Mock AIME 2024-2025 | — | 84% |
| Omni-MATH | — | 60.3% |
| FrontierMath (Feb 2025 set) | — | 19.7% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Grok 4: 53.8 (#55)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Expert | 1266 | 1415 |
| GPQA Diamond | — | 87% |
| MMLU-Pro | — | 85.1% |
| Confabulations | — | 12.4% |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, Grok 4: 33.7 (#94)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Vision | — | 1210 |
| GeoBench | — | 45% |
Multilingual Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Grok 4: 51.8 (#103)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Non-English | 1273 | 1403 |
| LMArena Chinese | 1318 | 1427 |
| LMArena French | 1289 | 1418 |
| LMArena German | 1258 | 1429 |
| LMArena Japanese | 1228 | 1394 |
| LMArena Korean | 1209 | 1377 |
| LMArena Russian | 1289 | 1410 |
| LMArena Spanish | 1248 | 1420 |
Instruction Following Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Grok 4: 79.2 (#5)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Instruction Following | 1280 | 1387 |
| IFEval | — | 94.9% |
Long Context Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Grok 4: 63.1 (#4)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Longer Query | 1301 | 1409 |
| Fiction.LiveBench | — | 94.4% |
Writing & Preference Grok 4 leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Grok 4: 58.5 (#116)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 |
|---|---|---|
| LMArena Text | 1294 | 1411 |
| LMArena Creative Writing | 1285 | 1397 |
| LMArena Multi-Turn | 1297 | 1416 |
| Short-Story Creative Writing | — | 76.9% |
| WildBench | — | 79.7% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Grok 4?
Grok 4 is the stronger model overall, scoring 48.1 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Grok 4 better for coding?
Grok 4 scores higher on coding benchmarks: 50.3 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Grok 4 share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Grok 4 has 48.