Model comparison
DeepSeek-V2.5 (Sep 2024) vs Grok-3 mini
Grok-3 mini is the stronger model overall, scoring 41.0 to 37.0 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 1 category and Grok-3 mini in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Grok-3 mini leads 46.3 to 34.2.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 49.3% for Grok-3 mini.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Grok-3 mini | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 37.0 | 41.0 |
| Released | 2024-09-06 | 2025-04-09 |
| Weights | Open | Proprietary |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 31.3 (#274), Grok-3 mini: 40.4 (#126)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| Aider Polyglot | 17.8% | 49.3% |
| LMArena Coding | 1310 | 1380 |
| WeirdML | — | 42.6% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 24.5 (#145), Grok-3 mini: 13.8 (#323)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1374 |
| ARC-AGI-2 | — | 0.4% |
| Kagi LLM Benchmark | — | 61.3% |
| ARC-AGI-1 | — | 16.5% |
| Epoch Capabilities Index | — | 140.35 |
Math Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 33.3 (#177), Grok-3 mini: 40.3 (#78)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Math | 1289 | 1386 |
| OTIS Mock AIME 2024-2025 | — | 77.8% |
| Omni-MATH | — | 31.8% |
| MATH Level 5 | — | 90.9% |
| FrontierMath (Feb 2025 set) | — | 5.9% |
Knowledge Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 34.2 (#190), Grok-3 mini: 46.3 (#75)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Expert | 1266 | 1395 |
| GPQA Diamond | — | 76.3% |
| MMLU-Pro | — | 79.9% |
| Confabulations | — | 10.8% |
| GPQA (HELM) | — | 67.5% |
Multilingual Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 42.7 (#188), Grok-3 mini: 48.3 (#141)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Non-English | 1273 | 1352 |
| LMArena Chinese | 1317 | 1387 |
| LMArena French | 1288 | 1357 |
| LMArena German | 1258 | 1350 |
| LMArena Japanese | 1228 | 1343 |
| LMArena Korean | 1209 | 1336 |
| LMArena Russian | 1289 | 1352 |
| LMArena Spanish | 1248 | 1379 |
Instruction Following Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 67.7 (#189), Grok-3 mini: 78.7 (#8)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Instruction Following | 1280 | 1356 |
| IFEval | — | 95.1% |
Long Context Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#169), Grok-3 mini: 41.0 (#142)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Longer Query | 1301 | 1372 |
| Fiction.LiveBench | — | 66.7% |
Writing & Preference Grok-3 mini leads
DeepSeek-V2.5 (Sep 2024): 50.0 (#182), Grok-3 mini: 52.7 (#163)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok-3 mini |
|---|---|---|
| LMArena Text | 1294 | 1370 |
| LMArena Creative Writing | 1285 | 1342 |
| LMArena Multi-Turn | 1297 | 1355 |
| Short-Story Creative Writing | — | 73.5% |
| WildBench | — | 65.1% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Grok-3 mini?
Grok-3 mini is the stronger model overall, scoring 41.0 to 37.0 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Grok-3 mini better for coding?
Grok-3 mini scores higher on coding benchmarks: 40.4 versus 31.3 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Grok-3 mini share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Grok-3 mini has 35.