Model comparison
DeepSeek V4 Flash vs GPT-3.5-turbo
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 23.1 on the Noometry Index.
Last verified . 27 shared benchmarks.
DeepSeekSummary
- They share 27 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 8 categories and GPT-3.5-turbo in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 6.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 94.4% for DeepSeek V4 Flash and 2.2% for GPT-3.5-turbo.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.50 / $1.50 for GPT-3.5-turbo.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 16K.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Flash | GPT-3.5-turbo | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 53.6 | 23.1 |
| Released | 2026-04-24 | 2023-03-01 |
| Weights | Open | Proprietary |
| Context window | 1M | 16K |
| Max output | 393K | 4K |
| Input $ / M tokens | $0.15 | $0.50 |
| Output $ / M tokens | $0.60 | $1.50 |
| Results tracked | 41 | 44 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
DeepSeek V4 Flash: 47.6 (#56), GPT-3.5-turbo: 23.5 (#326)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| WeirdML | 63% | 3.5% |
| LMArena Coding | 1457 | 1137 |
| FrontierCode | 18.8% | — |
| LMArena WebDev | 1581 | — |
| SciCode | 49.9% | — |
| BigCodeBench Instruct | — | 39.1% |
| BigCodeBench Complete | — | 50.6% |
| ALE-Bench | 1,306 | — |
| HumanEval+ | — | 70.7% |
| MBPP+ | — | 69.7% |
Agentic & Tool Use Not comparable
DeepSeek V4 Flash: —, GPT-3.5-turbo: —
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| METR Time Horizons | — | 21.5% |
Reasoning DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.7 (#25), GPT-3.5-turbo: 13.3 (#328)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| Chess Puzzles | 33% | 0% |
| LMArena Hard Prompts | 1443 | 1108 |
| Mystery Game Puzzles | 34% | 3% |
| DTBench | 90.9% | 48.5% |
| LMCA | 41.7% | 9.7% |
| Epoch Capabilities Index | 154.51 | 118.48 |
| ARC-AGI-2 | 61.4% | — |
| SimpleBench | 61.1% | — |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 89.6% | — |
| ARC-AGI-1 | 89% | — |
| CritPt | 16.6% | — |
| Adversarial NLI | — | 58.1% |
| BIG-Bench Hard | — | 61.6% |
| CommonsenseQA 2.0 | — | 57% |
| ForecastBench | — | 50.4 |
| WinoGrande | — | 81.6% |
Math DeepSeek V4 Flash leads
DeepSeek V4 Flash: 60.0 (#31), GPT-3.5-turbo: 6.0 (#321)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| FrontierMath (Tiers 1-3) | 57.5% | 0% |
| OTIS Mock AIME 2024-2025 | 94.4% | 2.2% |
| LMArena Math | 1427 | 1142 |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |
| MATH Level 5 | — | 15.9% |
| GSM8K | — | 57.8% |
Knowledge DeepSeek V4 Flash leads
DeepSeek V4 Flash: 55.5 (#43), GPT-3.5-turbo: 10.0 (#298)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| GPQA Diamond | 91% | 28% |
| LMArena Expert | 1441 | 1070 |
| SimpleQA Verified | 33.6% | — |
| ARC (AI2) Challenge | — | 87.4% |
| BoolQ | — | 87% |
| MMLU | — | 71.4% |
| OpenBookQA | — | 86% |
| TriviaQA | — | 85.8% |
Multilingual DeepSeek V4 Flash leads
DeepSeek V4 Flash: 53.2 (#71), GPT-3.5-turbo: 31.6 (#253)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| LMArena Non-English | 1420 | 1108 |
| LMArena Chinese | 1467 | 1075 |
| LMArena French | 1438 | 1117 |
| LMArena German | 1418 | 1090 |
| LMArena Japanese | 1403 | 1043 |
| LMArena Korean | 1383 | 1019 |
| LMArena Russian | 1428 | 1123 |
| LMArena Spanish | 1437 | 1120 |
Instruction Following DeepSeek V4 Flash leads
DeepSeek V4 Flash: 75.1 (#76), GPT-3.5-turbo: 58.1 (#257)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| LMArena Instruction Following | 1421 | 1119 |
Long Context DeepSeek V4 Flash leads
DeepSeek V4 Flash: 43.8 (#81), GPT-3.5-turbo: 34.0 (#249)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| LMArena Longer Query | 1435 | 1121 |
Writing & Preference DeepSeek V4 Flash leads
DeepSeek V4 Flash: 64.0 (#59), GPT-3.5-turbo: 25.4 (#300)
| Benchmark | DeepSeek V4 Flash | GPT-3.5-turbo |
|---|---|---|
| LMArena Text | 1432 | 1125 |
| LMArena Creative Writing | 1403 | 1092 |
| EQ-Bench Creative Writing | 1559 | 451 |
| LMArena Multi-Turn | 1449 | 1117 |
Frequently asked questions
Is DeepSeek V4 Flash better than GPT-3.5-turbo?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 23.1 on the Noometry Index.
Which is cheaper, DeepSeek V4 Flash or GPT-3.5-turbo?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-3.5-turbo lists at $0.50 and $1.50.
Is DeepSeek V4 Flash or GPT-3.5-turbo better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 23.5 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Flash does, with 1M tokens against 16K.
How many benchmarks do DeepSeek V4 Flash and GPT-3.5-turbo share?
27 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and GPT-3.5-turbo has 44.