Enter opens the highlighted result. Esc closes.

DeepSeek V3 vs R1 0528

Updated October 21, 2026. 10 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace DeepSeek V3 Replace R1 0528
DeepSeek V3DeepSeek
Coding
70.2
Overall
No data
Output $/M
$1.0287
0
R1 0528DeepSeek
Coding
71.4
Overall
13.1
Output $/M
$2.15
10
vs boards won of 10 shared

Decision reading

DeepSeek V3 and R1 0528 are level for coding, 70.2 against 71.4 on the Vibe Coding Index, so price and context decide.

Pick DeepSeek V3 when output tokens cost 2.1x less. Pick R1 0528 when what matters is Aider polyglot or GPQA Diamond (Epoch AI run), where it leads.

Model ADeepSeek V3DeepSeek70.2/1001 independent benchmarkNo cost per solved taskModel BR1 0528DeepSeek71.4/1001 independent benchmarkNo cost per solved task
Shared results
10
DeepSeek V3 only
3
R1 0528 only
49
Like-for-like categories
4 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.

At a glance

Coding index
DeepSeek V3 70.2 from 1 benchmark, R1 0528 71.4 from 1 benchmark.
Widest category
Math, 17.5 points apart on average over 1 shared benchmark.
Widest single gap
OTIS Mock AIME: DeepSeek V3 48.9% against R1 0528 66.4%, 17.5 points apart on a 0 to 100 scale.
Output price
DeepSeek V3 $1.0287 against R1 0528 $2.15 per million output tokens, 2.1x apart.
Context window
Both take 164K tokens.

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.

Boards won, where both have a result
CategoryDeepSeek V3BarsR1 0528Ties
Coding01
Reasoning01
Math01
Web design07
All shared boards010
CodingLike-for-likeFour of these boards feed the Vibe Coding Index01
Coding benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
Aider polyglot70.2%71.4%R1 0528
SWE-bench Lite30.7%Not measuredOnly DeepSeek V3 measured
LiveCodeBench (Artificial Analysis run)Not measured77.0%Only R1 0528 measured
Terminal-Bench Hard (Artificial Analysis run)Not measured15.9%Only R1 0528 measured
WeirdMLNot measured41.6%Only R1 0528 measured
ALE-BenchNot measured804.1Only R1 0528 measured
LMArena Text: codingNot measured1427Only R1 0528 measured
LMArena Text: software and ITNot measured1435Only R1 0528 measured
Neither model has a result on the other 30 boards here
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
Intelligence benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
AA Intelligence IndexNot measured13.1Only R1 0528 measured
Neither model has a result on the other 1 board here
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
Agentic work benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
τ²-Bench TelecomNot measured36.5%Only R1 0528 measured
METR time horizonsNot measured53.8%Only R1 0528 measured
DeepResearch BenchNot measured35.1%Only R1 0528 measured
Neither model has a result on the other 36 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
Reasoning benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
GPQA Diamond (Artificial Analysis run)Not measured81.3%Only R1 0528 measured
Humanity's Last Exam (Artificial Analysis run)Not measured15.8%Only R1 0528 measured
GPQA Diamond (Epoch AI run)71.2%76.3%R1 0528
ARC-AGINot measured21.2%Only R1 0528 measured
ARC-AGI-2Not measured1.1%Only R1 0528 measured
SimpleBenchNot measured40.8%Only R1 0528 measured
LMArena Text: hard promptsNot measured1415Only R1 0528 measured
LMArena Text: expertNot measured1394Only R1 0528 measured
LMArena Text: hard prompts in EnglishNot measured1422Only R1 0528 measured
Neither model has a result on the other 12 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index01
Math benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
AA Math IndexNot measured76.0Only R1 0528 measured
AIME 2025 (Artificial Analysis run)Not measured76.0%Only R1 0528 measured
OTIS Mock AIME48.9%66.4%R1 0528
MATH level 5Not measured96.6%Only R1 0528 measured
LMArena Text: mathNot measured1400Only R1 0528 measured
LMArena Text: mathematical workNot measured1415Only R1 0528 measured
Neither model has a result on the other 7 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
Knowledge benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
MMLU Pro (Artificial Analysis run)Not measured84.9%Only R1 0528 measured
LMArena Text: scienceNot measured1430Only R1 0528 measured
LMArena Text: medicine and healthcareNot measured1445Only R1 0528 measured
LMArena Text: legal and governmentNot measured1419Only R1 0528 measured
LMArena Text: business and financeNot measured1402Only R1 0528 measured
Neither model has a result on the other 12 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index07
Web design benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
Design Arena: websites11201149R1 0528
Design Arena: UI components10931102R1 0528
Design Arena: game dev10541098R1 0528
Design Arena: data viz10891182R1 0528
Design Arena: 3D10871111R1 0528
Design Arena: all code categories11081139R1 0528
Design Arena: SVG9761038R1 0528
Neither model has a result on the other 2 boards here
GeneralDirectional onlyPublic boards, shown as published, kept out of the index23 boards, none shared
General benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
IFBenchNot measured39.6%Only R1 0528 measured
Long context reasoning (AA-LCR)Not measured55.7%Only R1 0528 measured
DTBench62.7%Not measuredOnly DeepSeek V3 measured
LMCA15.2%Not measuredOnly DeepSeek V3 measured
Fiction.LiveBenchNot measured75.0%Only R1 0528 measured
Creative writing (Lech Mazur)Not measured81.9%Only R1 0528 measured
LMArena TextNot measured1428Only R1 0528 measured
LMArena Text: instruction followingNot measured1382Only R1 0528 measured
LMArena Text: creative writingNot measured1405Only R1 0528 measured
LMArena Text: EnglishNot measured1437Only R1 0528 measured
LMArena Text: non-EnglishNot measured1412Only R1 0528 measured
LMArena Text: ChineseNot measured1442Only R1 0528 measured
LMArena Text: GermanNot measured1404Only R1 0528 measured
LMArena Text: JapaneseNot measured1391Only R1 0528 measured
LMArena Text: KoreanNot measured1360Only R1 0528 measured
LMArena Text: FrenchNot measured1417Only R1 0528 measured
LMArena Text: SpanishNot measured1411Only R1 0528 measured
LMArena Text: RussianNot measured1423Only R1 0528 measured
LMArena Text: PolishNot measured1428Only R1 0528 measured
LMArena Text: multi-turnNot measured1405Only R1 0528 measured
LMArena Text: longer queriesNot measured1391Only R1 0528 measured
LMArena Text: writing and languageNot measured1402Only R1 0528 measured
LMArena Text: entertainment and mediaNot measured1400Only R1 0528 measured
Neither model has a result on the other 23 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for DeepSeek V3 and R1 0528
BenchmarkDeepSeek V3R1 0528Leader
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$0.0008Fits in one request
R1 0528$0.0016Fits in one request

DeepSeek V3 costs less

Repository review

50K input + 3K output tokens

DeepSeek V3$0.016Fits in one request
R1 0528$0.031Fits in one request

DeepSeek V3 costs less

Agent loop

200K input + 10K output tokens

DeepSeek V3$0.062Exceeds the context window
R1 0528$0.121Exceeds the context window

DeepSeek V3 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.

Specifications of DeepSeek V3 and R1 0528
SpecificationDeepSeek V3DeepSeek R1 0528
MakerDeepSeekDeepSeek
WeightsOpen weightsOpen weights
ReleasedDecember 26, 2024May 28, 2025
Context window164K164K
Input $/Mtok$0.2574$0.50
Output $/Mtok$1.0287$2.15
Coding index70.271.4
Independent measurements behind it11
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is DeepSeek V3 better than R1 0528 for coding?

DeepSeek V3 and R1 0528 are level for coding, 70.2 against 71.4 on the Vibe Coding Index, so price and context decide. Pick DeepSeek V3 when output tokens cost 2.1x less. Pick R1 0528 when what matters is Aider polyglot or GPQA Diamond (Epoch AI run), where it leads.

Which is cheaper, DeepSeek V3 or R1 0528?

DeepSeek V3 costs $0.2574 per million input tokens and $1.0287 per million output tokens; R1 0528 costs $0.50 and $2.15. Cost per solved task is known for neither of them.

Which has the bigger context window, DeepSeek V3 or R1 0528?

Neither: both take 164K tokens.

How many benchmarks is the DeepSeek V3 vs R1 0528 comparison based on?

10 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3's coding index rests on 1 independent benchmark and R1 0528's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next