DeepSeek V3 0324 vs DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 leads DeepSeek V3 0324 on the Artificial Analysis Intelligence Index, 30.4 to 9.7, and DeepSeek V3 0324 costs 2.8x less per output token ($1.07 vs $2.976 per million). Updated October 22, 2026; 14 benchmarks have a result for both.
DeepSeek V3 0324 or DeepSeek V4 Pro 0423: which to pick
Only DeepSeek V4 Pro 0423 has a coding score, 38.7 on the Vibe Coding Index; DeepSeek V3 0324 has none, so the two cannot be ranked on coding.
Pick DeepSeek V3 0324 for output tokens 2.8x cheaper ($1.07 vs $2.976 per million).
- Shared results
- 14
- DeepSeek V3 0324 only
- 21
- DeepSeek V4 Pro 0423 only
- 70
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
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, six public coding benchmarks
No call
Only DeepSeek V4 Pro 0423 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 leads, 30.4 against 9.7.
Same board, both measured
Agentic work
AA Agentic Index
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 leads, 26.3 against 0.8.
Same board, both measured
Reasoning
GPQA Diamond
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 leads, 89.4% against 61.6%.
Same board, both measured
Math
AIME
No call
Only DeepSeek V3 0324 has a AIME result, 52.2%.
Directional only
Knowledge
MMLU Pro
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 leads, 87.3% against 79.5%.
Same board, both measured
Web design
Design Arena: websites
No call
Only DeepSeek V4 Pro 0423 has a Design Arena: websites result, 1234.
Directional only
Long documents
Listed context window
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 takes 1.05M tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V3 0324
DeepSeek V3 0324 $0.0008, DeepSeek V4 Pro 0423 $0.0028 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only DeepSeek V4 Pro 0423 has the number, $0.57.
Directional only
At a glance
- Coding index
- DeepSeek V3 0324 not benchmarked, DeepSeek V4 Pro 0423 38.7 from 5 benchmarks.
- Widest category
- Math, 58.9 points apart on average over 1 shared benchmark.
- Widest single gap
- OTIS Mock AIME: DeepSeek V3 0324 37.8% against DeepSeek V4 Pro 0423 96.7%, 58.9 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3 0324 $1.07 against DeepSeek V4 Pro 0423 $2.976 per million output tokens, 2.8x apart.
- Context window
- DeepSeek V3 0324 takes 164K tokens, DeepSeek V4 Pro 0423 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 V3 0324 | Bars | DeepSeek V4 Pro 0423 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 2 | ||
| Math | 0 | 1 | ||
| Knowledge | 0 | 4 | ||
| General | 0 | 2 | ||
| All shared boards | 0 | 14 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| SWE-bench Verified | 42.0% | Not measured | Only DeepSeek V3 0324 measured |
| LMArena WebDev | Not measured | 1447 | Only DeepSeek V4 Pro 0423 measured |
| AA Coding Index | 21.2 | 59.4 | DeepSeek V4 Pro 0423 |
| LiveCodeBench (Vals AI) | 65.5% | 87.5% | DeepSeek V4 Pro 0423 |
| SWE-bench Verified (Vals AI) | Not measured | 77.4% | Only DeepSeek V4 Pro 0423 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 11.1% | Only DeepSeek V4 Pro 0423 measured |
| Vibe Code Bench (Vals AI) | Not measured | 49.9% | Only DeepSeek V4 Pro 0423 measured |
| Code migration (Vals AI) | Not measured | 26.2% | Only DeepSeek V4 Pro 0423 measured |
| ProgramBench (Vals AI) | Not measured | 0.0% | Only DeepSeek V4 Pro 0423 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 50.2% | Only DeepSeek V4 Pro 0423 measured |
| SkillsBench (Vals AI) | Not measured | 51.3% | Only DeepSeek V4 Pro 0423 measured |
| LiveCodeBench (Artificial Analysis run) | 40.5% | Not measured | Only DeepSeek V3 0324 measured |
| SciCode | 39.0% | Not measured | Only DeepSeek V3 0324 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 15.2% | Not measured | Only DeepSeek V3 0324 measured |
| FrontierCode | Not measured | 17.6% | Only DeepSeek V4 Pro 0423 measured |
| WeirdML | 36.1% | 48.9% | DeepSeek V4 Pro 0423 |
| ALE-Bench | Not measured | 1006.1 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: coding | Not measured | 1470 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: software and IT | Not measured | 1469 | Only DeepSeek V4 Pro 0423 measured |
| LMArena WebDev: React | Not measured | 1441 | Only DeepSeek V4 Pro 0423 measured |
| LMArena WebDev: HTML | Not measured | 1435 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 18 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| AA Intelligence Index | 9.7 | 30.4 | DeepSeek V4 Pro 0423 |
| Epoch Capabilities Index | Not measured | 149.1 | Only DeepSeek V4 Pro 0423 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| AA Agentic Index | 0.8 | 26.3 | DeepSeek V4 Pro 0423 |
| Design Arena: agents building web apps | Not measured | 1000 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: agents building full-stack apps | Not measured | 948 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: agents building Godot games | Not measured | 1059 | Only DeepSeek V4 Pro 0423 measured |
| Finance Agent (Vals AI) | Not measured | 44.1% | Only DeepSeek V4 Pro 0423 measured |
| Tax Agent Bench (Vals AI) | Not measured | 25.2% | Only DeepSeek V4 Pro 0423 measured |
| Legal Research Bench (Vals AI) | Not measured | 23.1% | Only DeepSeek V4 Pro 0423 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 51.6% | Only DeepSeek V4 Pro 0423 measured |
| τ²-Bench Telecom | 47.1% | Not measured | Only DeepSeek V3 0324 measured |
| METR time horizons | 49.6% | Not measured | Only DeepSeek V3 0324 measured |
| Vending-Bench 2 | Not measured | $3,285 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 30 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| GPQA Diamond | 61.6% | 89.4% | DeepSeek V4 Pro 0423 |
| GPQA Diamond (Artificial Analysis run) | 65.5% | Not measured | Only DeepSeek V3 0324 measured |
| Humanity's Last Exam (Artificial Analysis run) | 4.7% | Not measured | Only DeepSeek V3 0324 measured |
| GPQA Diamond (Epoch AI run) | 67.6% | 90.9% | DeepSeek V4 Pro 0423 |
| SimpleBench | 27.2% | Not measured | Only DeepSeek V3 0324 measured |
| Chess puzzles | Not measured | 19.0% | Only DeepSeek V4 Pro 0423 measured |
| Mystery game puzzles | Not measured | 17.0% | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: hard prompts | Not measured | 1461 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: expert | Not measured | 1463 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: hard prompts in English | Not measured | 1466 | Only DeepSeek V4 Pro 0423 measured |
| Kagi LLM Benchmark | Not measured | 53.5% | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 11 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| AIME | 52.2% | Not measured | Only DeepSeek V3 0324 measured |
| MATH 500 (Vals AI) | 88.6% | Not measured | Only DeepSeek V3 0324 measured |
| MGSM (Vals AI) | 91.7% | Not measured | Only DeepSeek V3 0324 measured |
| AA Math Index | 41.0 | Not measured | Only DeepSeek V3 0324 measured |
| AIME 2025 (Artificial Analysis run) | 41.0% | Not measured | Only DeepSeek V3 0324 measured |
| OTIS Mock AIME | 37.8% | 96.7% | DeepSeek V4 Pro 0423 |
| FrontierMath, tiers 1 to 3 | Not measured | 45.3% | Only DeepSeek V4 Pro 0423 measured |
| FrontierMath, tier 4 | Not measured | 2.4% | Only DeepSeek V4 Pro 0423 measured |
| ProofBench | Not measured | 16.0% | Only DeepSeek V4 Pro 0423 measured |
| MATH level 5 | 75.5% | Not measured | Only DeepSeek V3 0324 measured |
| LMArena Text: math | Not measured | 1443 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: mathematical work | Not measured | 1455 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 1 board here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| MMLU Pro | 79.5% | 87.3% | DeepSeek V4 Pro 0423 |
| LegalBench (Vals AI) | 77.7% | 80.3% | DeepSeek V4 Pro 0423 |
| CaseLaw (Vals AI) | Not measured | 59.4% | Only DeepSeek V4 Pro 0423 measured |
| CorpFin (Vals AI) | 54.7% | 61.4% | DeepSeek V4 Pro 0423 |
| TaxEval (Vals AI) | 71.1% | 72.1% | DeepSeek V4 Pro 0423 |
| MedQA (Vals AI) | 82.0% | Not measured | Only DeepSeek V3 0324 measured |
| MedCode (Vals AI) | Not measured | 40.5% | Only DeepSeek V4 Pro 0423 measured |
| MMLU Pro (Artificial Analysis run) | 81.9% | Not measured | Only DeepSeek V3 0324 measured |
| SimpleQA Verified | Not measured | 47.0% | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: science | Not measured | 1470 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: medicine and healthcare | Not measured | 1466 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: legal and government | Not measured | 1464 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: business and finance | Not measured | 1443 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1234 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: UI components | Not measured | 1219 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: game dev | Not measured | 1228 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: data viz | Not measured | 1203 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: 3D | Not measured | 1253 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: all code categories | Not measured | 1240 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: SVG | Not measured | 1139 | Only DeepSeek V4 Pro 0423 measured |
| Design Arena: ASCII art | Not measured | 1149 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| Public Benefits Bench (Vals AI) | Not measured | 62.9% | Only DeepSeek V4 Pro 0423 measured |
| MedScribe (Vals AI) | Not measured | 75.1% | Only DeepSeek V4 Pro 0423 measured |
| IFBench | 41.0% | Not measured | Only DeepSeek V3 0324 measured |
| Long context reasoning (AA-LCR) | 40.7% | Not measured | Only DeepSeek V3 0324 measured |
| DTBench | 64.8% | 90.7% | DeepSeek V4 Pro 0423 |
| LMCA | 15.5% | 41.2% | DeepSeek V4 Pro 0423 |
| Fiction.LiveBench | 50.0% | Not measured | Only DeepSeek V3 0324 measured |
| Creative writing (Lech Mazur) | 77.0% | Not measured | Only DeepSeek V3 0324 measured |
| Surface Evolver Bench | Not measured | 40.0% | Only DeepSeek V4 Pro 0423 measured |
| CL-bench Life | Not measured | 13.5% | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text | Not measured | 1451 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: instruction following | Not measured | 1445 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: creative writing | Not measured | 1446 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: English | Not measured | 1459 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: non-English | Not measured | 1439 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Chinese | Not measured | 1481 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: German | Not measured | 1458 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Japanese | Not measured | 1446 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Korean | Not measured | 1409 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: French | Not measured | 1473 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Spanish | Not measured | 1457 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Russian | Not measured | 1453 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: Polish | Not measured | 1461 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: multi-turn | Not measured | 1468 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: longer queries | Not measured | 1458 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: writing and language | Not measured | 1447 | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text: entertainment and media | Not measured | 1431 | Only DeepSeek V4 Pro 0423 measured |
| Neither model has a result on the other 19 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V3 0324 | DeepSeek V4 Pro 0423 | 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 V3 0324 costs less
Repository review
50K input + 3K output tokens
DeepSeek V3 0324 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V3 0324 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 V3 0324 | DeepSeek V4 Pro 0423 |
|---|---|---|
| Maker | DeepSeek | DeepSeek |
| Weights | Open weights | Open weights |
| Released | March 24, 2025 | April 24, 2026 |
| Context window | 164K | 1.05M |
| Input $/Mtok | $0.27 | $1.358 |
| Output $/Mtok | $1.07 | $2.976 |
| Coding index | Not benchmarked | 38.7 |
| Independent measurements behind it | 0 | 5 |
| Cost per solved task | Not computable | $0.57 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V3 0324 better than DeepSeek V4 Pro 0423 for coding?
Only DeepSeek V4 Pro 0423 has a coding score, 38.7 on the Vibe Coding Index; DeepSeek V3 0324 has none, so the two cannot be ranked on coding. Pick DeepSeek V3 0324 for output tokens 2.8x cheaper ($1.07 vs $2.976 per million).
Which is cheaper, DeepSeek V3 0324 or DeepSeek V4 Pro 0423?
DeepSeek V3 0324 costs $0.27 per million input tokens and $1.07 per million output tokens; DeepSeek V4 Pro 0423 costs $1.358 and $2.976. Cost per solved task is known for DeepSeek V4 Pro 0423 only.
Which has the bigger context window, DeepSeek V3 0324 or DeepSeek V4 Pro 0423?
DeepSeek V4 Pro 0423, with 1.05M tokens against 164K.
How many benchmarks is the DeepSeek V3 0324 vs DeepSeek V4 Pro 0423 comparison based on?
14 benchmarks have a result for both models as of October 22, 2026. DeepSeek V3 0324's coding index rests on 0 independent benchmarks and DeepSeek V4 Pro 0423's on 5. Where only one model has a result, the page says so instead of naming a winner.