Model comparison
DeepSeek-V3 vs GPT-5.4 mini
GPT-5.4 mini is the stronger model overall, scoring 44.8 to 39.2 on the Noometry Index. DeepSeek-V3 costs 4.2× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.
Last verified . 30 shared benchmarks.
DeepSeekSummary
- They share 30 benchmarks with published results for both. DeepSeek-V3 scores higher in 0 categories and GPT-5.4 mini in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 mini leads 45.4 to 30.4.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 88.9% for GPT-5.4 mini.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 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 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | GPT-5.4 mini | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 39.2 | 44.8 |
| Released | 2024-12-26 | 2026-03-17 |
| Weights | Open | Proprietary |
| Context window | 164K | 400K |
| Max output | 164K | 128K |
| Input $ / M tokens | $0.24 | $0.75 |
| Output $ / M tokens | $0.90 | $4.50 |
| Results tracked | 60 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 mini leads
DeepSeek-V3: 41.9 (#99), GPT-5.4 mini: 45.0 (#67)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| SciCode | 35.8% | 49.9% |
| WeirdML | 36.1% | 60.3% |
| LMArena Coding | 1368 | 1438 |
| FrontierCode | — | 27% |
| Aider Polyglot | 55.1% | — |
| LMArena WebDev | — | 1397 |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 1,189 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, GPT-5.4 mini: 29.6 (#74)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| DeepResearch Bench | — | 36.3% |
| METR Time Horizons | 49.6% | — |
Reasoning GPT-5.4 mini leads
DeepSeek-V3: 20.0 (#226), GPT-5.4 mini: 30.3 (#77)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 37.9% |
| CritPt | 0% | 10% |
| LMArena Hard Prompts | 1365 | 1424 |
| DTBench | 64.8% | 80% |
| LMCA | 15.5% | 40.8% |
| Epoch Capabilities Index | 135.95 | 148.83 |
| ForecastBench | 59.1 | 57 |
| ARC-AGI-2 | — | 18.9% |
| SimpleBench | 27.2% | — |
| NYT Connections (extended) | — | 61.8% |
| ARC-AGI-1 | — | 63.7% |
| Chess Puzzles | — | 24% |
| Thematic Generalization | — | 61.7% |
| LiveBench Reasoning | 65.8% | — |
| Mystery Game Puzzles | — | 11% |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | 87.5% | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math GPT-5.4 mini leads
DeepSeek-V3: 30.4 (#208), GPT-5.4 mini: 45.4 (#63)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 88.9% |
| LMArena Math | 1373 | 1418 |
| FrontierMath (Feb 2025 set) | 1.7% | 28.3% |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 9.8% |
| ProofBench | — | 21% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge GPT-5.4 mini leads
DeepSeek-V3: 37.5 (#147), GPT-5.4 mini: 51.4 (#62)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 67.6% | 86.9% |
| Vectara Hallucination Rate | 6.1% | 5.5% |
| LMArena Expert | 1352 | 1434 |
| SimpleQA Verified | — | 29.4% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, GPT-5.4 mini: 39.0 (#51)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | — | 1245 |
Multilingual GPT-5.4 mini leads
DeepSeek-V3: 48.8 (#138), GPT-5.4 mini: 52.1 (#94)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1358 | 1405 |
| LMArena Chinese | 1390 | 1446 |
| LMArena French | 1386 | 1440 |
| LMArena German | 1374 | 1407 |
| LMArena Japanese | 1333 | 1375 |
| LMArena Korean | 1321 | 1369 |
| LMArena Russian | 1372 | 1417 |
| LMArena Spanish | 1358 | 1403 |
Instruction Following GPT-5.4 mini leads
DeepSeek-V3: 73.0 (#125), GPT-5.4 mini: 74.3 (#99)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1345 | 1405 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context GPT-5.4 mini leads
DeepSeek-V3: 34.0 (#248), GPT-5.4 mini: 42.9 (#107)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1353 | 1407 |
| Fiction.LiveBench | 50% | — |
Writing & Preference GPT-5.4 mini leads
DeepSeek-V3: 57.6 (#124), GPT-5.4 mini: 64.2 (#56)
| Benchmark | DeepSeek-V3 | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1375 | 1411 |
| LMArena Creative Writing | 1364 | 1370 |
| EQ-Bench Creative Writing | 1472 | 1665 |
| LMArena Multi-Turn | 1389 | 1429 |
| Short-Story Creative Writing | 77% | — |
| WildBench | 83% | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than GPT-5.4 mini?
GPT-5.4 mini is the stronger model overall, scoring 44.8 to 39.2 on the Noometry Index. DeepSeek-V3 costs 4.2× less per token, which makes it the better buy when GPT-5.4 mini's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or GPT-5.4 mini?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.
Is DeepSeek-V3 or GPT-5.4 mini better for coding?
GPT-5.4 mini scores higher on coding benchmarks: 45.0 versus 41.9 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 and GPT-5.4 mini share?
30 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and GPT-5.4 mini has 46.