Enter opens the highlighted result. Esc closes.

DeepSeek V3.2 vs Kimi K2 Thinking

Kimi K2 Thinking leads DeepSeek V3.2 on the Artificial Analysis Intelligence Index, 22.0 to 21.5, and DeepSeek V3.2 costs 2.8x less per output token ($0.88 vs $2.50 per million). Updated October 21, 2026; 27 benchmarks have a result for both.

Replace DeepSeek V3.2 Replace Kimi K2 Thinking
DeepSeek V3.2DeepSeek
Coding
30.0
Overall
37.9
Output $/M
$0.88
15
Kimi K2 ThinkingMoonshot AI
Coding
No data
Overall
No data
Output $/M
$2.50
12
vs boards won of 27 shared

DeepSeek V3.2 or Kimi K2 Thinking: which to pick

Only DeepSeek V3.2 has a coding score, 30.0 on the Vibe Coding Index; Kimi K2 Thinking has none, so the two cannot be ranked on coding.

Pick Kimi K2 Thinking for a lower cost per solved task ($0.61 against $0.99) and a larger context window (262K against 164K tokens).

Model ADeepSeek V3.2DeepSeek30.0/1003 independent benchmarks$0.99 per solved taskModel BKimi K2 ThinkingMoonshot AINot benchmarked1 independent benchmark$0.61 per solved task
Shared results
27
DeepSeek V3.2 only
48
Kimi K2 Thinking only
3
Like-for-like categories
8 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.

At a glance

Coding index
DeepSeek V3.2 30.0 from 3 benchmarks, Kimi K2 Thinking not benchmarked.
Widest category
Web design, 7.8 points apart on average over 1 shared benchmark.
Widest single gap
LiveCodeBench (Vals AI): DeepSeek V3.2 80.7% against Kimi K2 Thinking 63.1%, 17.5 points apart on a 0 to 100 scale.
Output price
DeepSeek V3.2 $0.88 against Kimi K2 Thinking $2.50 per million output tokens, 2.8x apart.
Context window
DeepSeek V3.2 takes 164K tokens, Kimi K2 Thinking 262K.

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 V3.2BarsKimi K2 ThinkingTies
Coding60
Intelligence11
Agentic work01
Reasoning30
Math04
Knowledge34
Web design10
General12
All shared boards1512
CodingLike-for-likeSix of these boards feed the Vibe Coding Index60
Coding benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
SWE-bench Verified70.0%63.4%DeepSeek V3.2
SWE-bench Multilingual59.0%Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev1362Not measuredOnly DeepSeek V3.2 measured
AA Coding Index44.2Not measuredOnly DeepSeek V3.2 measured
LiveCodeBench (Vals AI)80.7%63.1%DeepSeek V3.2
SWE-bench Verified (Vals AI)67.6%60.2%DeepSeek V3.2
Vibe Code Bench (Vals AI)5.1%Not measuredOnly DeepSeek V3.2 measured
LiveCodeBench (Artificial Analysis run)86.2%85.3%DeepSeek V3.2
Terminal-Bench Hard (Artificial Analysis run)35.6%31.1%DeepSeek V3.2
Terminal-Bench39.6%35.7%DeepSeek V3.2
WeirdMLNot measured42.8%Only Kimi K2 Thinking measured
ALE-BenchNot measured597.5Only Kimi K2 Thinking measured
LMArena Text: coding1453Not measuredOnly DeepSeek V3.2 measured
LMArena Text: software and IT1449Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev: React1350Not measuredOnly DeepSeek V3.2 measured
LMArena WebDev: HTML1371Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 23 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index11
Intelligence benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
AA Intelligence Index21.522.0Kimi K2 Thinking
Epoch Capabilities Index146.1145.8DeepSeek V3.2
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
τ²-Bench Telecom90.6%93.0%Kimi K2 Thinking
APEX-Agents21.3%Not measuredOnly DeepSeek V3.2 measured
METR time horizonsNot measured59.2%Only Kimi K2 Thinking measured
Vending-Bench 2$1,034Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index30
Reasoning benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
GPQA Diamond80.3%78.5%DeepSeek V3.2
GPQA Diamond (Artificial Analysis run)84.0%83.8%DeepSeek V3.2
Humanity's Last Exam (Artificial Analysis run)24.6%23.8%DeepSeek V3.2
ARC-AGI57.0%Not measuredOnly DeepSeek V3.2 measured
ARC-AGI-24.0%Not measuredOnly DeepSeek V3.2 measured
LMArena Text: hard prompts1434Not measuredOnly DeepSeek V3.2 measured
LMArena Text: expert1436Not measuredOnly DeepSeek V3.2 measured
LMArena Text: hard prompts in English1446Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 14 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index04
Math benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
AIME84.6%85.4%Kimi K2 Thinking
MGSM (Vals AI)86.0%90.2%Kimi K2 Thinking
AA Math Index92.094.7Kimi K2 Thinking
AIME 2025 (Artificial Analysis run)92.0%94.7%Kimi K2 Thinking
LMArena Text: math1435Not measuredOnly DeepSeek V3.2 measured
LMArena Text: mathematical work1440Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index34
Knowledge benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
MMLU Pro84.9%81.1%DeepSeek V3.2
LegalBench (Vals AI)76.1%80.2%Kimi K2 Thinking
CaseLaw (Vals AI)55.4%65.7%Kimi K2 Thinking
CorpFin (Vals AI)51.0%60.6%Kimi K2 Thinking
TaxEval (Vals AI)68.2%71.7%Kimi K2 Thinking
MedQA (Vals AI)93.9%92.6%DeepSeek V3.2
MMLU Pro (Artificial Analysis run)86.2%84.8%DeepSeek V3.2
LMArena Text: science1446Not measuredOnly DeepSeek V3.2 measured
LMArena Text: medicine and healthcare1440Not measuredOnly DeepSeek V3.2 measured
LMArena Text: legal and government1425Not measuredOnly DeepSeek V3.2 measured
LMArena Text: business and finance1410Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 7 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index10
Web design benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
Design Arena: websites11751113DeepSeek V3.2
Design Arena: UI components1149Not measuredOnly DeepSeek V3.2 measured
Design Arena: game dev1132Not measuredOnly DeepSeek V3.2 measured
Design Arena: data viz1158Not measuredOnly DeepSeek V3.2 measured
Design Arena: 3D1133Not measuredOnly DeepSeek V3.2 measured
Design Arena: all code categories1164Not measuredOnly DeepSeek V3.2 measured
Design Arena: SVG1032Not measuredOnly DeepSeek V3.2 measured
Design Arena: ASCII art1083Not measuredOnly DeepSeek V3.2 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index12
General benchmarks for DeepSeek V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
IFBench60.7%68.1%Kimi K2 Thinking
Long context reasoning (AA-LCR)73.3%72.0%DeepSeek V3.2
DTBench85.6%Not measuredOnly DeepSeek V3.2 measured
LMCA28.8%Not measuredOnly DeepSeek V3.2 measured
CL-bench12.4%17.6%Kimi K2 Thinking
CL-bench Life7.4%Not measuredOnly DeepSeek V3.2 measured
LMArena Text1425Not measuredOnly DeepSeek V3.2 measured
LMArena Text: instruction following1413Not measuredOnly DeepSeek V3.2 measured
LMArena Text: creative writing1400Not measuredOnly DeepSeek V3.2 measured
LMArena Text: English1439Not measuredOnly DeepSeek V3.2 measured
LMArena Text: non-English1409Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Chinese1455Not measuredOnly DeepSeek V3.2 measured
LMArena Text: German1413Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Japanese1374Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Korean1370Not measuredOnly DeepSeek V3.2 measured
LMArena Text: French1429Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Spanish1424Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Russian1424Not measuredOnly DeepSeek V3.2 measured
LMArena Text: Polish1419Not measuredOnly DeepSeek V3.2 measured
LMArena Text: multi-turn1427Not measuredOnly DeepSeek V3.2 measured
LMArena Text: longer queries1428Not measuredOnly DeepSeek V3.2 measured
LMArena Text: writing and language1407Not measuredOnly DeepSeek V3.2 measured
LMArena Text: entertainment and media1395Not measuredOnly DeepSeek V3.2 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 V3.2 and Kimi K2 Thinking
BenchmarkDeepSeek V3.2Kimi K2 ThinkingLeader
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 V3.2$0.0007Fits in one request
Kimi K2 Thinking$0.0019Fits in one request

DeepSeek V3.2 costs less

Repository review

50K input + 3K output tokens

DeepSeek V3.2$0.017Fits in one request
Kimi K2 Thinking$0.037Fits in one request

DeepSeek V3.2 costs less

Agent loop

200K input + 10K output tokens

DeepSeek V3.2$0.067Exceeds the context window
Kimi K2 Thinking$0.145Fits in one request

DeepSeek V3.2 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 V3.2 and Kimi K2 Thinking
SpecificationDeepSeek V3.2MoonshotAI Kimi K2 Thinking
MakerDeepSeekMoonshot AI
WeightsOpen weightsOpen weights
ReleasedDecember 1, 2025November 6, 2025
Context window164K262K
Input $/Mtok$0.29$0.60
Output $/Mtok$0.88$2.50
Coding index30.0Not benchmarked
Independent measurements behind it31
Cost per solved task$0.99$0.61
SourceOpenRouter listingOpenRouter listing

Questions

Is DeepSeek V3.2 better than Kimi K2 Thinking for coding?

Only DeepSeek V3.2 has a coding score, 30.0 on the Vibe Coding Index; Kimi K2 Thinking has none, so the two cannot be ranked on coding. Pick Kimi K2 Thinking for a lower cost per solved task ($0.61 against $0.99) and a larger context window (262K against 164K tokens).

Which is cheaper, DeepSeek V3.2 or Kimi K2 Thinking?

DeepSeek V3.2 costs $0.29 per million input tokens and $0.88 per million output tokens; Kimi K2 Thinking costs $0.60 and $2.50. Per solved SWE-bench Verified task, DeepSeek V3.2 costs $0.99 and Kimi K2 Thinking $0.61.

Which has the bigger context window, DeepSeek V3.2 or Kimi K2 Thinking?

Kimi K2 Thinking, with 262K tokens against 164K.

How many benchmarks is the DeepSeek V3.2 vs Kimi K2 Thinking comparison based on?

27 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3.2's coding index rests on 3 independent benchmarks and Kimi K2 Thinking's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next