Model comparison
DeepSeek-V3.2-Exp vs GPT-5.4 mini
DeepSeek-V3.2-Exp and GPT-5.4 mini score almost the same on the Noometry Index (44.9 vs 44.8), so choose on price, context window or the category you care about most.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 6 categories and GPT-5.4 mini in 3 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 mini leads 30.3 to 22.0.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 61.8% for GPT-5.4 mini.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- GPT-5.4 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-5.4 mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.9 | 44.8 |
| Released | 2025-09-29 | 2026-03-17 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 66K | 128K |
| Input $ / M tokens | $0.26 | $0.75 |
| Output $ / M tokens | $0.38 | $4.50 |
| Results tracked | 48 | 46 |
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-5.4 mini: 45.0 (#67)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena WebDev | 1362 | 1397 |
| SciCode | 38.9% | 49.9% |
| WeirdML | 39.5% | 60.3% |
| LMArena Coding | 1454 | 1438 |
| FrontierCode | — | 27% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| ALE-Bench | — | 1,189 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 37.9 (#31), GPT-5.4 mini: 29.6 (#74)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| DeepResearch Bench | — | 36.3% |
| Vending-Bench 2 | 1,034 | — |
Reasoning GPT-5.4 mini leads
DeepSeek-V3.2-Exp: 22.0 (#185), GPT-5.4 mini: 30.3 (#77)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 4% | 18.9% |
| Kagi LLM Benchmark | 52.2% | 37.9% |
| NYT Connections (extended) | 36.7% | 61.8% |
| ARC-AGI-1 | 57% | 63.7% |
| CritPt | 2.9% | 10% |
| Chess Puzzles | 14% | 24% |
| Thematic Generalization | 65% | 61.7% |
| LMArena Hard Prompts | 1433 | 1424 |
| DTBench | 87.7% | 80% |
| LMCA | 29.1% | 40.8% |
| Epoch Capabilities Index | 146.28 | 148.83 |
| Mystery Game Puzzles | — | 11% |
| ForecastBench | — | 57 |
Math GPT-5.4 mini leads
DeepSeek-V3.2-Exp: 40.8 (#75), GPT-5.4 mini: 45.4 (#63)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 88.9% |
| ProofBench | 8% | 21% |
| LMArena Math | 1435 | 1418 |
| FrontierMath (Feb 2025 set) | 22.1% | 28.3% |
| FrontierMath Tier 4 (v1) | 2.1% | 2.1% |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 9.8% |
| MathArena Final-Answer Competitions | 57.7% | — |
Knowledge Too close to call
DeepSeek-V3.2-Exp: 51.6 (#61), GPT-5.4 mini: 51.4 (#62)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 83.4% | 86.9% |
| Vectara Hallucination Rate | 5.3% | 5.5% |
| LMArena Expert | 1436 | 1434 |
| SimpleQA Verified | — | 29.4% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, GPT-5.4 mini: 39.0 (#51)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | — | 1245 |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.4 (#85), GPT-5.4 mini: 52.1 (#94)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1409 | 1405 |
| LMArena Chinese | 1461 | 1446 |
| LMArena French | 1435 | 1440 |
| LMArena German | 1440 | 1407 |
| LMArena Japanese | 1377 | 1375 |
| LMArena Korean | 1372 | 1369 |
| LMArena Russian | 1423 | 1417 |
| LMArena Spanish | 1441 | 1403 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.7 (#89), GPT-5.4 mini: 74.3 (#99)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1413 | 1405 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), GPT-5.4 mini: 42.9 (#107)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1428 | 1407 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference GPT-5.4 mini leads
DeepSeek-V3.2-Exp: 62.5 (#73), GPT-5.4 mini: 64.2 (#56)
| Benchmark | DeepSeek-V3.2-Exp | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1425 | 1411 |
| LMArena Creative Writing | 1403 | 1370 |
| EQ-Bench Creative Writing | 1515 | 1665 |
| LMArena Multi-Turn | 1427 | 1429 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than GPT-5.4 mini?
DeepSeek-V3.2-Exp and GPT-5.4 mini score almost the same on the Noometry Index (44.9 vs 44.8), so choose on price, context window or the category you care about most.
Which is cheaper, DeepSeek-V3.2-Exp or GPT-5.4 mini?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.
Is DeepSeek-V3.2-Exp or GPT-5.4 mini better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 45.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 mini does, with 400K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and GPT-5.4 mini share?
37 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and GPT-5.4 mini has 46.