Model comparison
DeepSeek-V2.5 (Sep 2024) vs o4-mini
o4-mini is the stronger model overall, scoring 41.6 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 1 category and o4-mini in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in coding, where o4-mini leads 40.9 to 31.7.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 72% for o4-mini.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | o4-mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 37.6 | 41.6 |
| Released | 2024-09-06 | 2025-04-16 |
| Weights | Open | Proprietary |
| Context window | — | 200K |
| Max output | — | 100K |
| Input $ / M tokens | — | $1.10 |
| Output $ / M tokens | — | $4.40 |
| Results tracked | 22 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o4-mini leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), o4-mini: 40.9 (#127)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| Aider Polyglot | 17.8% | 72% |
| LMArena Coding | 1309 | 1368 |
| SWE-bench Verified (bash only) | — | 45% |
| GSO | — | 3.6% |
| WeirdML | — | 52.6% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| CadEval | — | 62% |
| ALE-Bench | — | 826.17 |
| AlgoTune | — | 1.72 |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, o4-mini: 32.6 (#61)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 53.2% |
| GDPval | — | 25.3% |
| METR Time Horizons | — | 63.9% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), o4-mini: 24.6 (#162)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1351 |
| ARC-AGI-2 | — | 6.1% |
| SimpleBench | — | 38.7% |
| Kagi LLM Benchmark | — | 67.6% |
| ARC-AGI-1 | — | 58.7% |
| CritPt | — | 0.6% |
| Chess Puzzles | — | 26% |
| EnigmaEval | — | 9.2% |
| Mystery Game Puzzles | — | 5% |
| DTBench | — | 77.6% |
| LMCA | — | 26.5% |
| Epoch Capabilities Index | — | 145.64 |
| ForecastBench | — | 61.8 |
Math o4-mini leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), o4-mini: 40.8 (#89)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Math | 1288 | 1389 |
| FrontierMath (Tiers 1-3) | — | 36.1% |
| FrontierMath Tier 4 | — | 4.9% |
| OTIS Mock AIME 2024-2025 | — | 81.7% |
| Omni-MATH | — | 72% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 24.8% |
| FrontierMath Tier 4 (v1) | — | 6.3% |
Knowledge o4-mini leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), o4-mini: 43.6 (#91)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Expert | 1266 | 1343 |
| GPQA Diamond | — | 79.6% |
| Humanity's Last Exam | — | 18.1% |
| SimpleQA Verified | — | 19.6% |
| MMLU-Pro | — | 82% |
| Confabulations | — | 15.8% |
| Vectara Hallucination Rate | — | 18.6% |
| GPQA (HELM) | — | 73.5% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, o4-mini: 40.2 (#49)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Vision | — | 1194 |
| GeoBench | — | 64% |
| VPCT | — | 57.5% |
Multilingual o4-mini leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), o4-mini: 47.0 (#154)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Non-English | 1273 | 1337 |
| LMArena Chinese | 1318 | 1354 |
| LMArena French | 1289 | 1364 |
| LMArena German | 1258 | 1336 |
| LMArena Japanese | 1228 | 1308 |
| LMArena Korean | 1209 | 1312 |
| LMArena Russian | 1289 | 1334 |
| LMArena Spanish | 1248 | 1347 |
Instruction Following o4-mini leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), o4-mini: 75.2 (#68)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Instruction Following | 1280 | 1321 |
| IFEval | — | 92.8% |
Long Context o4-mini leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), o4-mini: 45.5 (#33)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Longer Query | 1301 | 1315 |
| Fiction.LiveBench | — | 77.8% |
Writing & Preference o4-mini leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), o4-mini: 54.0 (#152)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | o4-mini |
|---|---|---|
| LMArena Text | 1294 | 1353 |
| LMArena Creative Writing | 1285 | 1294 |
| LMArena Multi-Turn | 1297 | 1350 |
| Short-Story Creative Writing | — | 75% |
| WildBench | — | 85.4% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than o4-mini?
o4-mini is the stronger model overall, scoring 41.6 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or o4-mini better for coding?
o4-mini scores higher on coding benchmarks: 40.9 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and o4-mini share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and o4-mini has 60.