Model comparison
DeepSeek-V3.2-Exp vs GPT-5
GPT-5 is the stronger model overall, scoring 50.9 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 12× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Last verified . 39 shared benchmarks.
DeepSeekSummary
- They share 39 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 2 categories and GPT-5 in 7 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 47.6.
- The biggest single-benchmark swing is Chess Puzzles: 14% for DeepSeek-V3.2-Exp and 37% for GPT-5.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $1.25 / $10 for GPT-5.
- GPT-5 accepts more context: 400K tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | GPT-5 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.3 | 50.9 |
| Released | 2025-09-29 | 2025-08-07 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $1.25 |
| Output $ / M tokens | $0.38 | $10 |
| Results tracked | 49 | 69 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
DeepSeek-V3.2-Exp: 46.5 (#65), GPT-5: 50.3 (#47)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 65% |
| Aider Polyglot | 74.2% | 88% |
| LMArena WebDev | 1362 | 1418 |
| SciCode | 38.9% | 42.9% |
| WeirdML | 39.5% | 60.7% |
| LMArena Coding | 1454 | 1436 |
| SWE-bench Verified | — | 73.6% |
| SWE-bench Multilingual | 59% | — |
| GSO | — | 6.9% |
| ALE-Bench | — | 1,162 |
| AlgoTune | — | 1.67 |
Agentic & Tool Use Too close to call
DeepSeek-V3.2-Exp: 32.7 (#59), GPT-5: 33.1 (#56)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| Terminal-Bench | 39.6% | 49.6% |
| APEX-Agents | 21.3% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| GDPval | — | 34.8% |
| Remote Labor Index | — | 1.7% |
| TheAgentCompany | 42.9% | — |
| DeepResearch Bench | — | 49.6% |
| BALROG | — | 32.8% |
| LMArena Search | — | 1133 |
| METR Time Horizons | — | 69.6% |
| Vending-Bench 2 | 1,034 | — |
Reasoning GPT-5 leads
DeepSeek-V3.2-Exp: 22.1 (#208), GPT-5: 38.3 (#64)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| ARC-AGI-2 | 4% | 9.9% |
| Kagi LLM Benchmark | 52.2% | 72.7% |
| ARC-AGI-1 | 57% | 65.7% |
| CritPt | 2.9% | 12.6% |
| Chess Puzzles | 14% | 37% |
| LMArena Hard Prompts | 1434 | 1416 |
| DTBench | 87.7% | 90.7% |
| LMCA | 29.1% | 40% |
| Epoch Capabilities Index | 146.27 | 150 |
| SimpleBench | — | 56.7% |
| NYT Connections (extended) | 36.7% | — |
| EnigmaEval | — | 10.5% |
| Thematic Generalization | 65% | — |
| EBR-Bench | — | 12.7% |
| Mystery Game Puzzles | — | 23% |
| ForecastBench | — | 61.4 |
Math GPT-5 leads
DeepSeek-V3.2-Exp: 41.7 (#87), GPT-5: 55.0 (#44)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 91.4% |
| ProofBench | 8% | 18% |
| LMArena Math | 1435 | 1407 |
| FrontierMath (Feb 2025 set) | 22.1% | 32.4% |
| FrontierMath Tier 4 (v1) | 2.1% | 12.5% |
| FrontierMath (Tiers 1-3) | — | 55.4% |
| FrontierMath Tier 4 | — | 22% |
| MathArena Final-Answer Competitions | 57.7% | — |
| Omni-MATH | — | 64.7% |
| MATH Level 5 | — | 98.1% |
Knowledge GPT-5 leads
DeepSeek-V3.2-Exp: 51.7 (#66), GPT-5: 56.6 (#43)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| GPQA Diamond | 83.4% | 86.2% |
| Vectara Hallucination Rate | 5.3% | 14.7% |
| LMArena Expert | 1436 | 1419 |
| Humanity's Last Exam | — | 25.3% |
| SimpleQA Verified | — | 50.1% |
| MMLU-Pro | — | 86.3% |
| Confabulations | — | 10.3% |
| GPQA (HELM) | — | 79.2% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-5: 46.8 (#13)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| LMArena Vision | — | 1232 |
| GeoBench | — | 81% |
| VPCT | — | 66% |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.2 (#90), GPT-5: 51.4 (#110)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| LMArena Non-English | 1409 | 1397 |
| LMArena Chinese | 1461 | 1422 |
| LMArena French | 1433 | 1410 |
| LMArena German | 1440 | 1416 |
| LMArena Japanese | 1374 | 1409 |
| LMArena Korean | 1371 | 1360 |
| LMArena Russian | 1424 | 1406 |
| LMArena Spanish | 1440 | 1399 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), GPT-5: 73.8 (#113)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1388 |
| IFEval | — | 87.5% |
Long Context GPT-5 leads
DeepSeek-V3.2-Exp: 47.6 (#16), GPT-5: 69.5 (#2)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| Fiction.LiveBench | 83.3% | 97.2% |
| LMArena Longer Query | 1428 | 1399 |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference GPT-5 leads
DeepSeek-V3.2-Exp: 62.4 (#77), GPT-5: 63.4 (#65)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 |
|---|---|---|
| LMArena Text | 1425 | 1406 |
| LMArena Creative Writing | 1403 | 1365 |
| EQ-Bench Creative Writing | 1515 | 1627 |
| LMArena Multi-Turn | 1427 | 1426 |
| Short-Story Creative Writing | — | 86% |
| WildBench | — | 85.7% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-5?
GPT-5 is the stronger model overall, scoring 50.9 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 12× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-5?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-5 lists at $1.25 and $10.
Is DeepSeek-V3.2-Exp or GPT-5 better for coding?
GPT-5 scores higher on coding benchmarks: 50.3 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-5 share?
39 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and GPT-5 has 69.