Model comparison
DeepSeek-V2.5 (Sep 2024) vs Grok 4 Fast
Grok 4 Fast is the stronger model overall, scoring 39.4 to 37.6 on the Noometry Index.
Last verified . 17 shared benchmarks.
DeepSeekSummary
- They share 17 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 2 categories and Grok 4 Fast in 6 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where Grok 4 Fast leads 63.2 to 39.5.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast | |
|---|---|---|
| Provider | DeepSeek | xAI |
| Noometry Index | 37.6 | 39.4 |
| Released | 2024-09-06 | 2025-09-19 |
| Weights | Open | Proprietary |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 30 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Grok 4 Fast: 32.5 (#271)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Coding | 1309 | 1429 |
| Aider Polyglot | 17.8% | — |
| LMArena WebDev | — | 1159 |
| WeirdML | — | 42.9% |
| 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 Fast: 29.5 (#86)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| τ²-bench Banking | — | 15.7% |
| Cybench | — | 30% |
| LMArena Search | — | 1171 |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Grok 4 Fast: 22.2 (#201)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1412 |
| ARC-AGI-2 | — | 5.3% |
| Kagi LLM Benchmark | — | 66.1% |
| ARC-AGI-1 | — | 48.5% |
| DTBench | — | 82.7% |
| Epoch Capabilities Index | — | 144.2 |
| ForecastBench | — | 60.5 |
Math Grok 4 Fast leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Grok 4 Fast: 38.9 (#123)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Math | 1288 | 1419 |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Grok 4 Fast: 32.0 (#214)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Expert | 1266 | 1411 |
| Vectara Hallucination Rate | — | 19.7% |
Multilingual Grok 4 Fast leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Grok 4 Fast: 51.3 (#111)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Non-English | 1273 | 1396 |
| LMArena Chinese | 1318 | 1457 |
| LMArena French | 1289 | 1431 |
| LMArena German | 1258 | 1383 |
| LMArena Japanese | 1228 | 1352 |
| LMArena Korean | 1209 | 1357 |
| LMArena Russian | 1289 | 1389 |
| LMArena Spanish | 1248 | 1415 |
Instruction Following Grok 4 Fast leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Grok 4 Fast: 73.2 (#121)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Instruction Following | 1280 | 1387 |
Long Context Grok 4 Fast leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Grok 4 Fast: 63.2 (#3)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Longer Query | 1301 | 1415 |
| Fiction.LiveBench | — | 94.4% |
Writing & Preference Grok 4 Fast leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Grok 4 Fast: 60.0 (#102)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Grok 4 Fast |
|---|---|---|
| LMArena Text | 1294 | 1407 |
| LMArena Creative Writing | 1285 | 1387 |
| LMArena Multi-Turn | 1297 | 1414 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Grok 4 Fast?
Grok 4 Fast is the stronger model overall, scoring 39.4 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Grok 4 Fast better for coding?
They score almost the same on coding (31.7 vs 32.5); test both on your own repository before choosing.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Grok 4 Fast share?
17 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Grok 4 Fast has 30.