Model comparison
DeepSeek-V3.2-Exp vs GPT-5 Mini
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 41.7 on the Noometry Index.
Last verified . 40 shared benchmarks.
DeepSeekSummary
- They share 40 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 6 categories and GPT-5 Mini in 3 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3.2-Exp leads 62.5 to 55.4.
- The biggest single-benchmark swing is SWE-bench Multilingual: 59% for DeepSeek-V3.2-Exp and 39.7% for GPT-5 Mini.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.25 / $2 for GPT-5 Mini.
- GPT-5 Mini 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 Mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.9 | 41.7 |
| Released | 2025-09-29 | 2025-08-07 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $0.25 |
| Output $ / M tokens | $0.38 | $2 |
| Results tracked | 48 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.4 (#61), GPT-5 Mini: 40.0 (#134)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 59.8% |
| SWE-bench Multilingual | 59% | 39.7% |
| SciCode | 38.9% | 39.2% |
| WeirdML | 39.5% | 52.7% |
| LMArena Coding | 1454 | 1406 |
| SWE-bench Verified | — | 64.7% |
| Aider Polyglot | 74.2% | — |
| LMArena WebDev | 1362 | — |
| ALE-Bench | — | 799.77 |
| AlgoTune | — | 1.38 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 37.9 (#31), GPT-5 Mini: 31.0 (#64)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| Terminal-Bench | 39.6% | 34.8% |
| Berkeley Function Calling Leaderboard | 56.7% | 55.5% |
| Vending-Bench 2 | 1,034 | -31.18 |
| TheAgentCompany | 42.9% | — |
Reasoning GPT-5 Mini leads
DeepSeek-V3.2-Exp: 22.0 (#185), GPT-5 Mini: 24.1 (#153)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| ARC-AGI-2 | 4% | 4.4% |
| Kagi LLM Benchmark | 52.2% | 70.3% |
| ARC-AGI-1 | 57% | 54.3% |
| CritPt | 2.9% | 0% |
| Chess Puzzles | 14% | 30% |
| LMArena Hard Prompts | 1433 | 1380 |
| DTBench | 87.7% | 80.5% |
| LMCA | 29.1% | 34.2% |
| Epoch Capabilities Index | 146.28 | 145.51 |
| NYT Connections (extended) | 36.7% | — |
| EnigmaEval | — | 8.2% |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 10% |
| ForecastBench | — | 61 |
Math GPT-5 Mini leads
DeepSeek-V3.2-Exp: 40.8 (#75), GPT-5 Mini: 46.3 (#59)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 86.7% |
| ProofBench | 8% | 9% |
| LMArena Math | 1435 | 1377 |
| FrontierMath (Feb 2025 set) | 22.1% | 27.2% |
| FrontierMath Tier 4 (v1) | 2.1% | 6.3% |
| FrontierMath (Tiers 1-3) | — | 46.7% |
| FrontierMath Tier 4 | — | 12.2% |
| MathArena Final-Answer Competitions | 57.7% | — |
| Omni-MATH | — | 72.2% |
| MATH Level 5 | — | 97.8% |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.6 (#61), GPT-5 Mini: 45.5 (#80)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| GPQA Diamond | 83.4% | 75% |
| Vectara Hallucination Rate | 5.3% | 12.9% |
| LMArena Expert | 1436 | 1377 |
| Humanity's Last Exam | — | 19.4% |
| SimpleQA Verified | — | 21.6% |
| MMLU-Pro | — | 83.5% |
| Confabulations | — | 13.3% |
| GPQA (HELM) | — | 75.6% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-5 Mini: 35.0 (#80)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| LMArena Vision | — | 1202 |
| VPCT | — | 40.2% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.4 (#85), GPT-5 Mini: 49.1 (#131)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| LMArena Non-English | 1409 | 1363 |
| LMArena Chinese | 1461 | 1387 |
| LMArena French | 1435 | 1385 |
| LMArena German | 1440 | 1366 |
| LMArena Japanese | 1377 | 1342 |
| LMArena Korean | 1372 | 1306 |
| LMArena Russian | 1423 | 1362 |
| LMArena Spanish | 1441 | 1355 |
Instruction Following GPT-5 Mini leads
DeepSeek-V3.2-Exp: 74.7 (#89), GPT-5 Mini: 76.4 (#43)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| LMArena Instruction Following | 1413 | 1357 |
| IFEval | — | 92.7% |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), GPT-5 Mini: 41.9 (#127)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| Fiction.LiveBench | 83.3% | 69.4% |
| LMArena Longer Query | 1428 | 1355 |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.5 (#73), GPT-5 Mini: 55.4 (#143)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5 Mini |
|---|---|---|
| LMArena Text | 1425 | 1373 |
| LMArena Creative Writing | 1403 | 1325 |
| EQ-Bench Creative Writing | 1515 | 1313 |
| LMArena Multi-Turn | 1427 | 1363 |
| Short-Story Creative Writing | — | 83.1% |
| WildBench | — | 85.5% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-5 Mini?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 41.7 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-5 Mini?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-5 Mini lists at $0.25 and $2.
Is DeepSeek-V3.2-Exp or GPT-5 Mini better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 40.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5 Mini does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-5 Mini share?
40 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and GPT-5 Mini has 60.