Model comparison
DeepSeek V4 Flash vs o1
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 40.9 on the Noometry Index.
Last verified . 28 shared benchmarks.
DeepSeekSummary
- They share 28 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 7 categories and o1 in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Flash leads 53.7 to 27.9.
- The biggest single-benchmark swing is ARC-AGI-1: 89% for DeepSeek V4 Flash and 30.7% for o1.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $15 / $60 for o1.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 200K.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Flash | o1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 53.6 | 40.9 |
| Released | 2026-04-24 | 2024-09-12 |
| Weights | Open | Proprietary |
| Context window | 1M | 200K |
| Max output | 393K | 100K |
| Input $ / M tokens | $0.15 | $15 |
| Output $ / M tokens | $0.60 | $60 |
| Results tracked | 41 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
DeepSeek V4 Flash: 47.9 (#59), o1: 46.1 (#70)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| WeirdML | 63% | 47.6% |
| LMArena Coding | 1457 | 1367 |
| FrontierCode | 18.8% | — |
| Aider Polyglot | — | 61.7% |
| LMArena WebDev | 1582 | — |
| SciCode | 49.9% | — |
| LiveBench Coding | — | 69.7% |
| CadEval | — | 56% |
| ALE-Bench | 1,306 | — |
| HumanEval+ | — | 89% |
| MBPP+ | — | 80.2% |
Agentic & Tool Use Not comparable
DeepSeek V4 Flash: —, o1: 24.6 (#117)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| Cybench | — | 10% |
| METR Time Horizons | — | 51.1% |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#30), o1: 27.9 (#111)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| SimpleBench | 61.1% | 41.7% |
| ARC-AGI-1 | 89% | 30.7% |
| Chess Puzzles | 33% | 15% |
| LMArena Hard Prompts | 1444 | 1371 |
| DTBench | 90.9% | 74.7% |
| LMCA | 41.7% | 22.3% |
| Epoch Capabilities Index | 154.49 | 141.91 |
| ARC-AGI-2 | 61.4% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| CritPt | 16.6% | — |
| EnigmaEval | — | 5.7% |
| LiveBench Reasoning | — | 91.6% |
| Mystery Game Puzzles | 34% | — |
| LiveBench Data Analysis | — | 65.5% |
| LiveBench | — | 75.7% |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.3 (#37), o1: 36.1 (#175)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 57.5% | 14.7% |
| OTIS Mock AIME 2024-2025 | 94.4% | 73.3% |
| LMArena Math | 1427 | 1388 |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |
| LiveBench Math | — | 80.3% |
| MATH Level 5 | — | 94.7% |
| FrontierMath (Feb 2025 set) | — | 9.3% |
Knowledge DeepSeek V4 Flash leads
DeepSeek V4 Flash: 55.4 (#48), o1: 41.5 (#110)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| GPQA Diamond | 91% | 76.8% |
| SimpleQA Verified | 33.6% | 41.1% |
| LMArena Expert | 1441 | 1361 |
| Humanity's Last Exam | — | 8% |
| Confabulations | — | 11.7% |
Multimodal Not comparable
DeepSeek V4 Flash: —, o1: 34.2 (#93)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| LMArena Vision | — | 1168 |
| GeoBench | — | 80% |
| VPCT | — | 37% |
| SpatialViz-Bench | — | 41.4% |
Multilingual DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.0 (#72), o1: 48.6 (#142)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| LMArena Non-English | 1420 | 1358 |
| LMArena Chinese | 1468 | 1394 |
| LMArena French | 1439 | 1344 |
| LMArena German | 1418 | 1337 |
| LMArena Japanese | 1406 | 1346 |
| LMArena Korean | 1384 | 1396 |
| LMArena Russian | 1428 | 1356 |
| LMArena Spanish | 1436 | 1345 |
Instruction Following Too close to call
DeepSeek V4 Flash: 74.9 (#81), o1: 74.8 (#86)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| LMArena Instruction Following | 1421 | 1367 |
| LiveBench Instruction Following | — | 81.5% |
Long Context o1 leads
DeepSeek V4 Flash: 43.8 (#85), o1: 50.3 (#9)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| LMArena Longer Query | 1434 | 1378 |
| Fiction.LiveBench | — | 83.3% |
Writing & Preference DeepSeek V4 Flash leads
DeepSeek V4 Flash: 63.8 (#61), o1: 55.6 (#144)
| Benchmark | DeepSeek V4 Flash | o1 |
|---|---|---|
| LMArena Text | 1432 | 1366 |
| LMArena Creative Writing | 1403 | 1348 |
| LMArena Multi-Turn | 1449 | 1369 |
| Short-Story Creative Writing | — | 70.2% |
| EQ-Bench Creative Writing | 1559 | — |
| LiveBench Language | — | 65.4% |
Frequently asked questions
Is DeepSeek V4 Flash better than o1?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 40.9 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or o1?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; o1 lists at $15 and $60.
Is DeepSeek V4 Flash or o1 better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.9 versus 46.1 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 200K.
How many benchmarks do DeepSeek V4 Flash and o1 share?
28 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and o1 has 52.