DeepSeek V4.1 Flash vs GPT-5.4
Updated October 21, 2026. 75 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick DeepSeek V4.1 Flash for coding: it scores 77.2 on the Vibe Coding Index against 37.7 for GPT-5.4.
Pick GPT-5.4 when the context window is larger, 1.05M against 1.05M tokens, or when what matters is ALE-Bench or AA Intelligence Index, where it leads.
- Shared results
- 75
- DeepSeek V4.1 Flash only
- 26
- GPT-5.4 only
- 34
- Like-for-like categories
- 8 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, four public coding benchmarks
DeepSeek V4.1 Flash
DeepSeek V4.1 Flash scores 77.2 against 37.7.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-5.4
GPT-5.4 leads, 39.0 against 24.7.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only GPT-5.4 has a GPQA Diamond result, 91.7%.
Directional only
Math
AIME
No call
Only GPT-5.4 has a AIME result, 96.7%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.4 has a MMLU Pro result, 87.5%.
Directional only
Web design
Design Arena: websites
DeepSeek V4.1 Flash
DeepSeek V4.1 Flash leads, 1309 against 1199.
Same board, both measured
Long documents
Listed context window
GPT-5.4
GPT-5.4 takes 1.05M tokens against 1.05M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4.1 Flash
DeepSeek V4.1 Flash $0.0003, GPT-5.4 $0.010 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.4 has the number, $1.02.
Directional only
At a glance
- Coding index
- DeepSeek V4.1 Flash 77.2 from 2 benchmarks, GPT-5.4 37.7 from 1 benchmark.
- Widest category
- Web design, 16.4 points apart on average over 8 shared benchmarks.
- Widest single gap
- Design Arena: 3D: DeepSeek V4.1 Flash 1356 against GPT-5.4 1065, 36.4 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V4.1 Flash $0.525 against GPT-5.4 $15.00 per million output tokens, 28.6x apart.
- Context window
- DeepSeek V4.1 Flash takes 1.05M tokens, GPT-5.4 1.05M.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | DeepSeek V4.1 Flash | Bars | GPT-5.4 | Ties |
|---|---|---|---|---|
| Coding | 9 | 1 | 1 tie | |
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 3 | 7 | 1 tie | |
| Math | 1 | 5 | ||
| Knowledge | 1 | 6 | ||
| Web design | 8 | 0 | ||
| General | 8 | 21 | ||
| All shared boards | 30 | 43 | 2 ties |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index91
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| LiveBench coding | 90.2% | Not measured | Only DeepSeek V4.1 Flash measured |
| LMArena WebDev | 1618 | 1464 | DeepSeek V4.1 Flash |
| AA Coding Index | Not measured | 71.1 | Only GPT-5.4 measured |
| LiveCodeBench (Vals AI) | Not measured | 84.1% | Only GPT-5.4 measured |
| SWE-bench Verified (Vals AI) | Not measured | 78.2% | Only GPT-5.4 measured |
| Terminal-Bench 4 (Vals AI) | 19.7% | Not measured | Only DeepSeek V4.1 Flash measured |
| Vibe Code Bench (Vals AI) | 84.7% | 67.4% | DeepSeek V4.1 Flash |
| Code migration (Vals AI) | 45.6% | 35.0% | DeepSeek V4.1 Flash |
| ProgramBench (Vals AI) | 0.5% | 0.5% | Tie |
| IOI (Vals AI) | 40.3% | Not measured | Only DeepSeek V4.1 Flash measured |
| Terminal-Bench 2.1 (Vals AI) | 74.5% | Not measured | Only DeepSeek V4.1 Flash measured |
| SWE-bench Multimodal (Vals AI) | 46.5% | Not measured | Only DeepSeek V4.1 Flash measured |
| Vibe Code Bench 1-100 (Vals AI) | 16.4% | Not measured | Only DeepSeek V4.1 Flash measured |
| SkillsBench (Vals AI) | 69.8% | 51.7% | DeepSeek V4.1 Flash |
| CyberBench (Vals AI) | 73.7% | Not measured | Only DeepSeek V4.1 Flash measured |
| SciCode | 51.9% | Not measured | Only DeepSeek V4.1 Flash measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 57.6% | Only GPT-5.4 measured |
| Terminal-Bench | Not measured | 81.8% | Only GPT-5.4 measured |
| FrontierCode | 43.9% | Not measured | Only DeepSeek V4.1 Flash measured |
| GSO-Bench | Not measured | 31.4% | Only GPT-5.4 measured |
| WeirdML | Not measured | 77.7% | Only GPT-5.4 measured |
| MirrorCode | Not measured | 15.6% | Only GPT-5.4 measured |
| ALE-Bench | 1092.3 | 1607.0 | GPT-5.4 |
| LMArena Text: coding | 1502 | 1496 | DeepSeek V4.1 Flash |
| LMArena Text: software and IT | 1496 | 1491 | DeepSeek V4.1 Flash |
| LMArena WebDev: React | 1628 | 1454 | DeepSeek V4.1 Flash |
| LMArena WebDev: HTML | 1612 | 1478 | DeepSeek V4.1 Flash |
| LMArena WebDev: from an image | 1606 | 1436 | DeepSeek V4.1 Flash |
| Neither model has a result on the other 10 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| AA Intelligence Index | 24.7 | 39.0 | GPT-5.4 |
| Epoch Capabilities Index | 155.9 | 157.3 | GPT-5.4 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | Not measured | 1059 | Only GPT-5.4 measured |
| Design Arena: agents building full-stack apps | Not measured | 1017 | Only GPT-5.4 measured |
| Design Arena: agents building mobile apps | Not measured | 1087 | Only GPT-5.4 measured |
| Design Arena: agents building Android apps | Not measured | 1037 | Only GPT-5.4 measured |
| Design Arena: agents building Godot games | Not measured | 1135 | Only GPT-5.4 measured |
| LiveBench research | 86.8% | Not measured | Only DeepSeek V4.1 Flash measured |
| LiveBench browser | 84.0% | Not measured | Only DeepSeek V4.1 Flash measured |
| LiveBench long agentic | 40.1% | Not measured | Only DeepSeek V4.1 Flash measured |
| SREBench (Vals AI) | 0.8% | Not measured | Only DeepSeek V4.1 Flash measured |
| Finance Agent (Vals AI) | 53.5% | Not measured | Only DeepSeek V4.1 Flash measured |
| Tax Agent Bench (Vals AI) | 26.6% | Not measured | Only DeepSeek V4.1 Flash measured |
| Legal Research Bench (Vals AI) | 41.4% | Not measured | Only DeepSeek V4.1 Flash measured |
| Excel Modeling Benchmark (Vals AI) | 57.2% | Not measured | Only DeepSeek V4.1 Flash measured |
| BioMysteryBench (Vals AI) | 67.8% | Not measured | Only DeepSeek V4.1 Flash measured |
| Terminal-Bench-Science (Vals AI) | 0.0% | Not measured | Only DeepSeek V4.1 Flash measured |
| Vals Science Index | 29.6% | Not measured | Only DeepSeek V4.1 Flash measured |
| τ²-Bench Telecom | Not measured | 87.1% | Only GPT-5.4 measured |
| APEX-Agents | 39.5% | 52.4% | GPT-5.4 |
| METR time horizons | Not measured | 74.3% | Only GPT-5.4 measured |
| DeepResearch Bench | Not measured | 35.1% | Only GPT-5.4 measured |
| PostTrainBench | Not measured | 19.0% | Only GPT-5.4 measured |
| Vending-Bench 2 | Not measured | $6,144 | Only GPT-5.4 measured |
| Neither model has a result on the other 17 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index37
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 91.7% | Only GPT-5.4 measured |
| LiveBench reasoning | 86.7% | Not measured | Only DeepSeek V4.1 Flash measured |
| MysteryMechanism (Vals AI) | 21.2% | Not measured | Only DeepSeek V4.1 Flash measured |
| MMMU Pro (Vals AI) | Not measured | 87.5% | Only GPT-5.4 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 92.0% | Only GPT-5.4 measured |
| Humanity's Last Exam (Artificial Analysis run) | 39.2% | 43.7% | GPT-5.4 |
| GPQA Diamond (Epoch AI run) | 89.8% | 93.3% | GPT-5.4 |
| ARC-AGI | 94.5% | 93.7% | DeepSeek V4.1 Flash |
| ARC-AGI-2 | 72.9% | 74.0% | GPT-5.4 |
| Chess puzzles | 48.0% | 48.0% | Tie |
| Mystery game puzzles | 43.0% | 37.0% | DeepSeek V4.1 Flash |
| EBR-bench | Not measured | 25.4% | Only GPT-5.4 measured |
| LMArena Text: hard prompts | 1483 | 1485 | GPT-5.4 |
| LMArena Text: expert | 1501 | 1507 | GPT-5.4 |
| LMArena Text: hard prompts in English | 1488 | 1479 | DeepSeek V4.1 Flash |
| LMArena Vision: diagrams | 1297 | 1327 | GPT-5.4 |
| LMArena Vision: homework | 1307 | 1336 | GPT-5.4 |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index15
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| AIME | Not measured | 96.7% | Only GPT-5.4 measured |
| OTIS Mock AIME | 98.3% | 97.8% | DeepSeek V4.1 Flash |
| FrontierMath, tiers 1 to 3 | 67.4% | 78.6% | GPT-5.4 |
| FrontierMath, tier 4 | 26.8% | 49.0% | GPT-5.4 |
| ProofBench | 54.0% | 56.0% | GPT-5.4 |
| LMArena Text: math | 1478 | 1488 | GPT-5.4 |
| LMArena Text: mathematical work | 1485 | 1497 | GPT-5.4 |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index16
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 87.5% | Only GPT-5.4 measured |
| LegalBench (Vals AI) | 83.3% | 86.0% | GPT-5.4 |
| CaseLaw (Vals AI) | Not measured | 63.8% | Only GPT-5.4 measured |
| CorpFin (Vals AI) | Not measured | 65.3% | Only GPT-5.4 measured |
| TaxEval (Vals AI) | Not measured | 74.0% | Only GPT-5.4 measured |
| MortgageTax (Vals AI) | Not measured | 68.3% | Only GPT-5.4 measured |
| MedQA (Vals AI) | Not measured | 96.1% | Only GPT-5.4 measured |
| MedCode (Vals AI) | 41.2% | 41.3% | GPT-5.4 |
| SimpleQA Verified | 42.9% | 45.1% | GPT-5.4 |
| LMArena Text: science | 1469 | 1479 | GPT-5.4 |
| LMArena Text: medicine and healthcare | 1484 | 1458 | DeepSeek V4.1 Flash |
| LMArena Text: legal and government | 1443 | 1486 | GPT-5.4 |
| LMArena Text: business and finance | 1461 | 1474 | GPT-5.4 |
| Neither model has a result on the other 4 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index80
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| Design Arena: websites | 1309 | 1199 | DeepSeek V4.1 Flash |
| Design Arena: UI components | 1314 | 1215 | DeepSeek V4.1 Flash |
| Design Arena: game dev | 1331 | 1195 | DeepSeek V4.1 Flash |
| Design Arena: data viz | 1290 | 1196 | DeepSeek V4.1 Flash |
| Design Arena: 3D | 1356 | 1065 | DeepSeek V4.1 Flash |
| Design Arena: all code categories | 1320 | 1180 | DeepSeek V4.1 Flash |
| Design Arena: SVG | 1243 | 1141 | DeepSeek V4.1 Flash |
| Design Arena: ASCII art | 1265 | 1188 | DeepSeek V4.1 Flash |
| LiveBench design | 79.2% | Not measured | Only DeepSeek V4.1 Flash measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index821
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| LiveBench data analysis | 79.3% | Not measured | Only DeepSeek V4.1 Flash measured |
| Public Benefits Bench (Vals AI) | 64.3% | Not measured | Only DeepSeek V4.1 Flash measured |
| SAGE (Vals AI) | 47.9% | 43.3% | DeepSeek V4.1 Flash |
| MedScribe (Vals AI) | 85.5% | 77.5% | DeepSeek V4.1 Flash |
| IFBench | Not measured | 73.9% | Only GPT-5.4 measured |
| Long context reasoning (AA-LCR) | 84.0% | 82.0% | DeepSeek V4.1 Flash |
| DTBench | 89.9% | 94.4% | GPT-5.4 |
| LMCA | 47.0% | 52.0% | GPT-5.4 |
| CL-bench | Not measured | 27.9% | Only GPT-5.4 measured |
| Surface Evolver Bench | 46.3% | Not measured | Only DeepSeek V4.1 Flash measured |
| Furniture assembly | 34.2% | 37.5% | GPT-5.4 |
| CL-bench Life | Not measured | 21.7% | Only GPT-5.4 measured |
| LMArena Text | 1464 | 1469 | GPT-5.4 |
| LMArena Text: instruction following | 1473 | 1469 | DeepSeek V4.1 Flash |
| LMArena Text: creative writing | 1437 | 1439 | GPT-5.4 |
| LMArena Vision | 1278 | 1303 | GPT-5.4 |
| LMArena Text: English | 1473 | 1467 | DeepSeek V4.1 Flash |
| LMArena Text: non-English | 1449 | 1465 | GPT-5.4 |
| LMArena Text: Chinese | 1501 | 1520 | GPT-5.4 |
| LMArena Text: German | 1470 | 1472 | GPT-5.4 |
| LMArena Text: Japanese | 1432 | 1485 | GPT-5.4 |
| LMArena Text: Korean | 1436 | 1449 | GPT-5.4 |
| LMArena Text: French | 1465 | 1493 | GPT-5.4 |
| LMArena Text: Spanish | 1459 | 1452 | DeepSeek V4.1 Flash |
| LMArena Text: Russian | 1467 | 1480 | GPT-5.4 |
| LMArena Text: Polish | 1457 | 1472 | GPT-5.4 |
| LMArena Text: multi-turn | 1461 | 1482 | GPT-5.4 |
| LMArena Text: longer queries | 1473 | 1473 | DeepSeek V4.1 Flash |
| LMArena Text: writing and language | 1450 | 1458 | GPT-5.4 |
| LMArena Text: entertainment and media | 1445 | 1438 | DeepSeek V4.1 Flash |
| LMArena Vision: English | 1288 | 1301 | GPT-5.4 |
| LMArena Vision: Chinese | 1327 | 1354 | GPT-5.4 |
| LMArena Vision: reading text | 1282 | 1312 | GPT-5.4 |
| LMArena Vision: humor | 1260 | 1284 | GPT-5.4 |
| LMArena Vision: recognizing things | Not measured | 1251 | Only GPT-5.4 measured |
| LMArena Vision: creative writing | 1275 | 1283 | GPT-5.4 |
| LMArena Document | Not measured | 1471 | Only GPT-5.4 measured |
| Neither model has a result on the other 9 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V4.1 Flash | GPT-5.4 | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
DeepSeek V4.1 Flash costs less
Repository review
50K input + 3K output tokens
DeepSeek V4.1 Flash costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4.1 Flash costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | DeepSeek V4.1 Flash | OpenAI GPT-5.4 |
|---|---|---|
| Maker | DeepSeek | OpenAI |
| Weights | Open weights | No public weights listed |
| Released | September 10, 2026 | March 5, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.049 | $2.50 |
| Output $/Mtok | $0.525 | $15.00 |
| Coding index | 77.2 | 37.7 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | Not computable | $1.02 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V4.1 Flash better than GPT-5.4 for coding?
Pick DeepSeek V4.1 Flash for coding: it scores 77.2 on the Vibe Coding Index against 37.7 for GPT-5.4. Pick GPT-5.4 when the context window is larger, 1.05M against 1.05M tokens, or when what matters is ALE-Bench or AA Intelligence Index, where it leads.
Which is cheaper, DeepSeek V4.1 Flash or GPT-5.4?
DeepSeek V4.1 Flash costs $0.049 per million input tokens and $0.525 per million output tokens; GPT-5.4 costs $2.50 and $15.00. Cost per solved task is known for GPT-5.4 only.
Which has the bigger context window, DeepSeek V4.1 Flash or GPT-5.4?
GPT-5.4, with 1.05M tokens against 1.05M.
How many benchmarks is the DeepSeek V4.1 Flash vs GPT-5.4 comparison based on?
75 benchmarks have a result for both models as of October 21, 2026. DeepSeek V4.1 Flash's coding index rests on 2 independent benchmarks and GPT-5.4's on 1. Where only one model has a result, the page says so instead of naming a winner.