Model comparison
DeepSeek-V3.2-Exp vs GPT-4.1 mini
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 33.6 on the Noometry Index.
Last verified . 35 shared benchmarks.
DeepSeekSummary
- They share 35 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 8 categories and GPT-4.1 mini in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3.2-Exp leads 41.7 to 24.1.
- The biggest single-benchmark swing is ARC-AGI-1: 57% for DeepSeek-V3.2-Exp and 3.5% for GPT-4.1 mini.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.40 / $1.60 for GPT-4.1 mini.
- GPT-4.1 mini accepts more context: 1.05M tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | GPT-4.1 mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.3 | 33.6 |
| Released | 2025-09-29 | 2025-04-14 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 66K | 33K |
| Input $ / M tokens | $0.26 | $0.40 |
| Output $ / M tokens | $0.38 | $1.60 |
| Results tracked | 49 | 47 |
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: 46.5 (#65), GPT-4.1 mini: 30.6 (#293)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 23.9% |
| Aider Polyglot | 74.2% | 32.4% |
| SciCode | 38.9% | 40.4% |
| WeirdML | 39.5% | 37.6% |
| LMArena Coding | 1454 | 1367 |
| LMArena WebDev | 1362 | — |
| SWE-bench Multilingual | 59% | — |
| BigCodeBench Instruct | — | 48.9% |
| CadEval | — | 16% |
Agentic & Tool Use Too close to call
DeepSeek-V3.2-Exp: 32.7 (#59), GPT-4.1 mini: 33.3 (#55)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| Berkeley Function Calling Leaderboard | 56.7% | 50.5% |
| Terminal-Bench | 39.6% | — |
| APEX-Agents | 21.3% | — |
| TheAgentCompany | 42.9% | — |
| Vending-Bench 2 | 1,034 | — |
Reasoning DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 22.1 (#208), GPT-4.1 mini: 10.8 (#340)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| ARC-AGI-2 | 4% | 0% |
| Kagi LLM Benchmark | 52.2% | 48.6% |
| ARC-AGI-1 | 57% | 3.5% |
| CritPt | 2.9% | 0% |
| Chess Puzzles | 14% | 7% |
| LMArena Hard Prompts | 1434 | 1349 |
| DTBench | 87.7% | 68.8% |
| LMCA | 29.1% | 21.1% |
| Epoch Capabilities Index | 146.27 | 135.01 |
| NYT Connections (extended) | 36.7% | — |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 7% |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 41.7 (#87), GPT-4.1 mini: 24.1 (#270)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 44.7% |
| LMArena Math | 1435 | 1343 |
| FrontierMath (Feb 2025 set) | 22.1% | 4.5% |
| FrontierMath (Tiers 1-3) | — | 6.7% |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| Omni-MATH | — | 49.1% |
| MATH Level 5 | — | 87.3% |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.7 (#66), GPT-4.1 mini: 34.7 (#194)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| GPQA Diamond | 83.4% | 65.8% |
| LMArena Expert | 1436 | 1338 |
| SimpleQA Verified | — | 12.7% |
| MMLU-Pro | — | 78.3% |
| Vectara Hallucination Rate | 5.3% | — |
| GPQA (HELM) | — | 61.4% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-4.1 mini: 35.8 (#82)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| LMArena Vision | — | 1181 |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 52.2 (#90), GPT-4.1 mini: 45.7 (#166)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| LMArena Non-English | 1409 | 1318 |
| LMArena Chinese | 1461 | 1329 |
| LMArena French | 1433 | 1358 |
| LMArena German | 1440 | 1351 |
| LMArena Japanese | 1374 | 1290 |
| LMArena Korean | 1371 | 1298 |
| LMArena Russian | 1424 | 1324 |
| LMArena Spanish | 1440 | 1319 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), GPT-4.1 mini: 73.7 (#118)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| LMArena Instruction Following | 1413 | 1333 |
| IFEval | — | 90.4% |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), GPT-4.1 mini: 31.8 (#275)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| Fiction.LiveBench | 83.3% | 44.4% |
| LMArena Longer Query | 1428 | 1344 |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.4 (#77), GPT-4.1 mini: 48.6 (#199)
| Benchmark | DeepSeek-V3.2-Exp | GPT-4.1 mini |
|---|---|---|
| LMArena Text | 1425 | 1340 |
| LMArena Creative Writing | 1403 | 1300 |
| EQ-Bench Creative Writing | 1515 | 1147 |
| LMArena Multi-Turn | 1427 | 1354 |
| WildBench | — | 83.8% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-4.1 mini?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 33.6 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-4.1 mini?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-4.1 mini lists at $0.40 and $1.60.
Is DeepSeek-V3.2-Exp or GPT-4.1 mini better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 30.6 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 mini does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-4.1 mini share?
35 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and GPT-4.1 mini has 47.