Model comparison
DeepSeek-V2.5 (Sep 2024) vs Grok 3
Grok 3 is the stronger model overall, scoring 39.9 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 2 categories and Grok 3 in 6 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek-V2.5 (Sep 2024) leads 25.6 to 13.7.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 53.3% for Grok 3.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Grok 3 | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 37.6 | 39.9 |
| Released | 2024-09-06 | 2025-04-09 |
| Weights | Open | Proprietary |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 40 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Grok 3: 41.9 (#115)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| Aider Polyglot | 17.8% | 53.3% |
| LMArena Coding | 1309 | 1432 |
| WeirdML | — | 37.2% |
| 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 3: 30.5 (#76)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| BALROG | — | 29.5% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Grok 3: 13.7 (#333)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1434 |
| ARC-AGI-2 | — | 0% |
| SimpleBench | — | 36.1% |
| Kagi LLM Benchmark | — | 61.3% |
| ARC-AGI-1 | — | 5.5% |
| Epoch Capabilities Index | — | 138.33 |
Math Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Grok 3: 38.0 (#145)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Math | 1288 | 1391 |
| OTIS Mock AIME 2024-2025 | — | 55.6% |
| Omni-MATH | — | 46.4% |
| MATH Level 5 | — | 88.7% |
| FrontierMath (Feb 2025 set) | — | 3.8% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Grok 3: 46.2 (#82)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Expert | 1266 | 1421 |
| GPQA Diamond | — | 75.8% |
| MMLU-Pro | — | 78.8% |
| Confabulations | — | 14.2% |
| Vectara Hallucination Rate | — | 5.8% |
| GPQA (HELM) | — | 65% |
Multilingual Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Grok 3: 52.3 (#87)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Non-English | 1273 | 1410 |
| LMArena Chinese | 1318 | 1448 |
| LMArena French | 1289 | 1460 |
| LMArena German | 1258 | 1431 |
| LMArena Japanese | 1228 | 1387 |
| LMArena Korean | 1209 | 1373 |
| LMArena Russian | 1289 | 1416 |
| LMArena Spanish | 1248 | 1417 |
Instruction Following Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Grok 3: 75.0 (#73)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Instruction Following | 1280 | 1409 |
| IFEval | — | 88.4% |
Long Context Too close to call
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Grok 3: 38.7 (#192)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Longer Query | 1301 | 1439 |
| Fiction.LiveBench | — | 58.3% |
Writing & Preference Grok 3 leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Grok 3: 55.8 (#141)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 3 |
|---|---|---|
| LMArena Text | 1294 | 1426 |
| LMArena Creative Writing | 1285 | 1414 |
| LMArena Multi-Turn | 1297 | 1425 |
| Short-Story Creative Writing | — | 76.4% |
| EQ-Bench Creative Writing | — | 1186 |
| WildBench | — | 84.9% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Grok 3?
Grok 3 is the stronger model overall, scoring 39.9 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Grok 3 better for coding?
Grok 3 scores higher on coding benchmarks: 41.9 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Grok 3 share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Grok 3 has 40.