Enter opens the highlighted result. Esc closes.

Kimi K2.6 vs Qwen3.5 397B A17B

Kimi K2.6 leads Qwen3.5 397B A17B on the Artificial Analysis Intelligence Index, 27.0 to 21.4, and Qwen3.5 397B A17B costs 1.7x less per output token ($2.34 vs $4.00 per million). Updated October 22, 2026; 62 benchmarks have a result for both.

Replace Kimi K2.6 Replace Qwen3.5 397B A17B
Kimi K2.6Moonshot AI
Coding
52.1
Overall
60.5
Output $/M
$4.00
58
Qwen3.5 397B A17BQwen
Coding
No data
Overall
57.1
Output $/M
$2.34
3
vs boards won of 62 shared, 1 tied

Kimi K2.6 or Qwen3.5 397B A17B: which to pick

Only Kimi K2.6 has a coding score, 52.1 on the Vibe Coding Index; Qwen3.5 397B A17B has none, so the two cannot be ranked on coding.

Pick Qwen3.5 397B A17B for output tokens 1.7x cheaper ($2.34 vs $4.00 per million) and work measured by IFBench or LMCA, where it leads.

Model AKimi K2.6Moonshot AI52.1/1003 independent benchmarks$0.35 per solved taskModel BQwen3.5 397B A17BQwenNot benchmarked1 independent benchmarkNo cost per solved task
Shared results
62
Kimi K2.6 only
47
Qwen3.5 397B A17B 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
Kimi K2.6 52.1 from 3 benchmarks, Qwen3.5 397B A17B not benchmarked.
Widest category
Web design, 10.8 points apart on average over 7 shared benchmarks.
Widest single gap
FrontierMath, tiers 1 to 3: Kimi K2.6 57.2% against Qwen3.5 397B A17B 31.2%, 26.0 points apart on a 0 to 100 scale.
Output price
Kimi K2.6 $4.00 against Qwen3.5 397B A17B $2.34 per million output tokens, 1.7x apart.
Context window
Both take 262K 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
CategoryKimi K2.6BarsQwen3.5 397B A17BTies
Coding80
Intelligence20
Agentic work10
Reasoning801 tie
Math40
Knowledge40
Web design70
General243
All shared boards5831 tie
CodingLike-for-likeSix of these boards feed the Vibe Coding Index80
Coding benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
LMArena WebDev15091400Kimi K2.6
AA Coding Index61.8Not measuredOnly Kimi K2.6 measured
LiveCodeBench (Vals AI)86.8%Not measuredOnly Kimi K2.6 measured
SWE-bench Verified (Vals AI)76.2%Not measuredOnly Kimi K2.6 measured
Vibe Code Bench (Vals AI)37.9%Not measuredOnly Kimi K2.6 measured
Code migration (Vals AI)27.8%Not measuredOnly Kimi K2.6 measured
ProgramBench (Vals AI)0.0%Not measuredOnly Kimi K2.6 measured
Terminal-Bench 2.1 (Vals AI)53.6%Not measuredOnly Kimi K2.6 measured
SciCode51.5%44.8%Kimi K2.6
Terminal-Bench Hard (Artificial Analysis run)43.9%40.9%Kimi K2.6
WeirdML55.9%Not measuredOnly Kimi K2.6 measured
ALE-Bench1092.7621.9Kimi K2.6
LMArena Text: coding14891464Kimi K2.6
LMArena Text: software and IT14841463Kimi K2.6
LMArena WebDev: React15051388Kimi K2.6
LMArena WebDev: HTML14851410Kimi K2.6
LMArena WebDev: from an image1521Not measuredOnly Kimi K2.6 measured
Neither model has a result on the other 22 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
AA Intelligence Index27.021.4Kimi K2.6
Epoch Capabilities Index150.9146.8Kimi K2.6
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
Agentic work benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
AA Agentic Index20.5Not measuredOnly Kimi K2.6 measured
Design Arena: agents building web apps1268Not measuredOnly Kimi K2.6 measured
Design Arena: agents building full-stack apps1155Not measuredOnly Kimi K2.6 measured
Design Arena: agents building mobile apps1156Not measuredOnly Kimi K2.6 measured
Design Arena: agents building Android apps1161Not measuredOnly Kimi K2.6 measured
Design Arena: agents building Godot games1120Not measuredOnly Kimi K2.6 measured
Design Arena: agents building games1141Not measuredOnly Kimi K2.6 measured
Design Arena: HTML slides1135Not measuredOnly Kimi K2.6 measured
Design Arena: PowerPoint slides1105Not measuredOnly Kimi K2.6 measured
Design Arena: spreadsheets1276Not measuredOnly Kimi K2.6 measured
Finance Agent (Vals AI)44.9%Not measuredOnly Kimi K2.6 measured
Tax Agent Bench (Vals AI)12.4%Not measuredOnly Kimi K2.6 measured
Legal Research Bench (Vals AI)15.9%Not measuredOnly Kimi K2.6 measured
Excel Modeling Benchmark (Vals AI)57.9%Not measuredOnly Kimi K2.6 measured
τ²-Bench Telecom95.9%95.6%Kimi K2.6
APEX-AgentsNot measured24.9%Only Qwen3.5 397B A17B measured
OSWorld 2.04.6%Not measuredOnly Kimi K2.6 measured
Vending-Bench 2$6,205Not measuredOnly Kimi K2.6 measured
Neither model has a result on the other 23 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index80
Reasoning benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
GPQA Diamond89.1%Not measuredOnly Kimi K2.6 measured
MMMU Pro (Vals AI)86.3%Not measuredOnly Kimi K2.6 measured
GPQA Diamond (Artificial Analysis run)91.1%89.3%Kimi K2.6
Humanity's Last Exam (Artificial Analysis run)37.5%29.0%Kimi K2.6
GPQA Diamond (Epoch AI run)90.8%86.4%Kimi K2.6
Chess puzzles20.0%Not measuredOnly Kimi K2.6 measured
Mystery game puzzles18.0%18.0%Tie
EBR-bench2.4%Not measuredOnly Kimi K2.6 measured
LMArena Text: hard prompts14701448Kimi K2.6
LMArena Text: expert14911462Kimi K2.6
LMArena Text: hard prompts in English14711450Kimi K2.6
LMArena Vision: diagrams12981277Kimi K2.6
LMArena Vision: homework13101292Kimi K2.6
Neither model has a result on the other 9 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index40
Math benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
OTIS Mock AIME96.1%88.9%Kimi K2.6
FrontierMath, tiers 1 to 357.2%31.2%Kimi K2.6
FrontierMath, tier 425.6%Not measuredOnly Kimi K2.6 measured
ProofBench16.0%Not measuredOnly Kimi K2.6 measured
LMArena Text: math14741453Kimi K2.6
LMArena Text: mathematical work14801454Kimi K2.6
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
Knowledge benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
MMLU Pro87.6%Not measuredOnly Kimi K2.6 measured
LegalBench (Vals AI)84.7%Not measuredOnly Kimi K2.6 measured
CaseLaw (Vals AI)61.2%Not measuredOnly Kimi K2.6 measured
CorpFin (Vals AI)66.7%Not measuredOnly Kimi K2.6 measured
TaxEval (Vals AI)74.7%Not measuredOnly Kimi K2.6 measured
MortgageTax (Vals AI)65.8%Not measuredOnly Kimi K2.6 measured
MedCode (Vals AI)40.1%Not measuredOnly Kimi K2.6 measured
SimpleQA Verified34.9%Not measuredOnly Kimi K2.6 measured
LMArena Text: science14681457Kimi K2.6
LMArena Text: medicine and healthcare14571455Kimi K2.6
LMArena Text: legal and government14621441Kimi K2.6
LMArena Text: business and finance14451435Kimi K2.6
Neither model has a result on the other 6 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index70
Web design benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
Design Arena: websites12691190Kimi K2.6
Design Arena: UI components12641161Kimi K2.6
Design Arena: game dev12501140Kimi K2.6
Design Arena: data viz12491175Kimi K2.6
Design Arena: 3D12671160Kimi K2.6
Design Arena: all code categories12711179Kimi K2.6
Design Arena: SVG11801139Kimi K2.6
Design Arena: ASCII art1165Not measuredOnly Kimi K2.6 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index243
General benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
Public Benefits Bench (Vals AI)56.6%Not measuredOnly Kimi K2.6 measured
SAGE (Vals AI)50.2%Not measuredOnly Kimi K2.6 measured
MedScribe (Vals AI)78.2%Not measuredOnly Kimi K2.6 measured
Vals Multimodal Index56.4%Not measuredOnly Kimi K2.6 measured
IFBench76.0%78.8%Qwen3.5 397B A17B
Long context reasoning (AA-LCR)81.0%77.3%Kimi K2.6
DTBench90.9%87.5%Kimi K2.6
LMCA37.3%37.9%Qwen3.5 397B A17B
Furniture assembly21.7%Not measuredOnly Kimi K2.6 measured
ExploitBench18.4%Not measuredOnly Kimi K2.6 measured
LMArena Text14551437Kimi K2.6
LMArena Text: instruction following14511424Kimi K2.6
LMArena Text: creative writing14341402Kimi K2.6
LMArena Vision12821262Kimi K2.6
LMArena Text: English14601443Kimi K2.6
LMArena Text: non-English14461430Kimi K2.6
LMArena Text: Chinese15201498Kimi K2.6
LMArena Text: German14511445Kimi K2.6
LMArena Text: Japanese14431429Kimi K2.6
LMArena Text: Korean14271384Kimi K2.6
LMArena Text: French14711462Kimi K2.6
LMArena Text: Spanish14641441Kimi K2.6
LMArena Text: Russian14461430Kimi K2.6
LMArena Text: Polish14381447Qwen3.5 397B A17B
LMArena Text: multi-turn14531446Kimi K2.6
LMArena Text: longer queries14681441Kimi K2.6
LMArena Text: writing and language14391415Kimi K2.6
LMArena Text: entertainment and media14281399Kimi K2.6
LMArena Vision: English12871268Kimi K2.6
LMArena Vision: Chinese13461314Kimi K2.6
LMArena Vision: reading text12911271Kimi K2.6
LMArena Vision: humor12801262Kimi K2.6
LMArena Vision: captioningNot measured1230Only Qwen3.5 397B A17B measured
LMArena Vision: recognizing thingsNot measured1265Only Qwen3.5 397B A17B measured
LMArena Vision: creative writing12731260Kimi K2.6
LMArena Document1451Not measuredOnly Kimi K2.6 measured
Neither model has a result on the other 10 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Kimi K2.6 and Qwen3.5 397B A17B
BenchmarkKimi K2.6Qwen3.5 397B A17BLeader
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

Kimi K2.6$0.0029Fits in one request
Qwen3.5 397B A17B$0.0016Fits in one request

Qwen3.5 397B A17B costs less

Repository review

50K input + 3K output tokens

Kimi K2.6$0.059Fits in one request
Qwen3.5 397B A17B$0.027Fits in one request

Qwen3.5 397B A17B costs less

Agent loop

200K input + 10K output tokens

Kimi K2.6$0.230Fits in one request
Qwen3.5 397B A17B$0.101Fits in one request

Qwen3.5 397B A17B 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 Kimi K2.6 and Qwen3.5 397B A17B
SpecificationMoonshotAI Kimi K2.6Qwen3.5 397B A17B
MakerMoonshot AIQwen
WeightsOpen weightsOpen weights
ReleasedApril 20, 2026February 16, 2026
Context window262K262K
Input $/Mtok$0.95$0.39
Output $/Mtok$4.00$2.34
Coding index52.1Not benchmarked
Independent measurements behind it31
Cost per solved task$0.35Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Kimi K2.6 better than Qwen3.5 397B A17B for coding?

Only Kimi K2.6 has a coding score, 52.1 on the Vibe Coding Index; Qwen3.5 397B A17B has none, so the two cannot be ranked on coding. Pick Qwen3.5 397B A17B for output tokens 1.7x cheaper ($2.34 vs $4.00 per million) and work measured by IFBench or LMCA, where it leads.

Which is cheaper, Kimi K2.6 or Qwen3.5 397B A17B?

Kimi K2.6 costs $0.95 per million input tokens and $4.00 per million output tokens; Qwen3.5 397B A17B costs $0.39 and $2.34. Cost per solved task is known for Kimi K2.6 only.

Which has the bigger context window, Kimi K2.6 or Qwen3.5 397B A17B?

Neither: both take 262K tokens.

How many benchmarks is the Kimi K2.6 vs Qwen3.5 397B A17B comparison based on?

62 benchmarks have a result for both models as of October 22, 2026. Kimi K2.6's coding index rests on 3 independent benchmarks and Qwen3.5 397B A17B's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next