Enter opens the highlighted result. Esc closes.

Kimi K2.5 vs Qwen3.8 27B

Updated October 21, 2026. 64 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Kimi K2.5 Replace Qwen3.8 27B
Kimi K2.5Moonshot AI
Coding
58.5
Overall
23.5
Output $/M
$2.50
26
Qwen3.8 27BQwen
Coding
73.6
Overall
33.7
Output $/M
$1.35
38
vs boards won of 64 shared

Decision reading

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 58.5 for Kimi K2.5.

Pick Kimi K2.5 when a solved task costs $0.24 against $1.26, or when what matters is LMArena Text: coding or MMMU Pro (Vals AI), where it leads.

Model AKimi K2.5Moonshot AI58.5/1002 independent benchmarks$0.24 per solved taskModel BQwen3.8 27BQwen73.6/1002 independent benchmarks$1.26 per solved task
Shared results
64
Kimi K2.5 only
34
Qwen3.8 27B only
21
Like-for-like categories
7 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
Kimi K2.5 58.5 from 2 benchmarks, Qwen3.8 27B 73.6 from 2 benchmarks.
Widest category
Agentic work, 21.4 points apart on average over 3 shared benchmarks.
Widest single gap
Vibe Code Bench (Vals AI): Kimi K2.5 17.5% against Qwen3.8 27B 64.8%, 47.3 points apart on a 0 to 100 scale.
Output price
Kimi K2.5 $2.50 against Qwen3.8 27B $1.35 per million output tokens, 1.9x apart.
Context window
Kimi K2.5 takes 262K tokens, Qwen3.8 27B 1M.

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 K2.5BarsQwen3.8 27BTies
Coding111
Intelligence02
Agentic work03
Reasoning29
Math20
Knowledge53
General1610
All shared boards2638
CodingLike-for-likeFour of these boards feed the Vibe Coding Index111
Coding benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
SWE-bench Verified70.8%Not measuredOnly Kimi K2.5 measured
SWE-bench Multilingual67.3%Not measuredOnly Kimi K2.5 measured
LiveBench codingNot measured86.8%Only Qwen3.8 27B measured
LMArena WebDev14371593Qwen3.8 27B
AA Coding Index46.868.1Qwen3.8 27B
LiveCodeBench (Vals AI)83.9%84.0%Qwen3.8 27B
SWE-bench Verified (Vals AI)70.0%86.0%Qwen3.8 27B
Vibe Code Bench (Vals AI)17.5%64.8%Qwen3.8 27B
Code migration (Vals AI)7.0%14.2%Qwen3.8 27B
ProgramBench (Vals AI)Not measured0.0%Only Qwen3.8 27B measured
IOI (Vals AI)Not measured39.1%Only Qwen3.8 27B measured
Terminal-Bench 2.1 (Vals AI)42.0%58.4%Qwen3.8 27B
SkillsBench (Vals AI)Not measured38.1%Only Qwen3.8 27B measured
CyberBench (Vals AI)Not measured67.8%Only Qwen3.8 27B measured
SciCodeNot measured46.6%Only Qwen3.8 27B measured
Terminal-Bench Hard (Artificial Analysis run)34.8%Not measuredOnly Kimi K2.5 measured
Terminal-Bench43.2%Not measuredOnly Kimi K2.5 measured
WeirdML45.6%Not measuredOnly Kimi K2.5 measured
ALE-Bench821.6Not measuredOnly Kimi K2.5 measured
LMArena Text: coding14831483Kimi K2.5
LMArena Text: software and IT14711478Qwen3.8 27B
LMArena WebDev: React14291602Qwen3.8 27B
LMArena WebDev: HTML14261570Qwen3.8 27B
LMArena WebDev: from an image14391577Qwen3.8 27B
Neither model has a result on the other 14 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
AA Intelligence Index23.533.7Qwen3.8 27B
Epoch Capabilities Index148.1151.7Qwen3.8 27B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index03
Agentic work benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
AA Agentic IndexNot measured45.8Only Qwen3.8 27B measured
Design Arena: agents building web apps1136Not measuredOnly Kimi K2.5 measured
Design Arena: agents building full-stack apps1114Not measuredOnly Kimi K2.5 measured
Design Arena: agents building mobile apps1147Not measuredOnly Kimi K2.5 measured
Design Arena: agents building Android apps1124Not measuredOnly Kimi K2.5 measured
Design Arena: agents building Godot games1219Not measuredOnly Kimi K2.5 measured
LiveBench researchNot measured59.4%Only Qwen3.8 27B measured
LiveBench browserNot measured32.0%Only Qwen3.8 27B measured
LiveBench long agenticNot measured27.9%Only Qwen3.8 27B measured
Finance Agent (Vals AI)35.8%48.5%Qwen3.8 27B
Tax Agent Bench (Vals AI)Not measured30.9%Only Qwen3.8 27B measured
Legal Research Bench (Vals AI)15.9%36.1%Qwen3.8 27B
Excel Modeling Benchmark (Vals AI)28.5%59.7%Qwen3.8 27B
Terminal-Bench-Science (Vals AI)Not measured0.0%Only Qwen3.8 27B measured
τ²-Bench Telecom95.9%Not measuredOnly Kimi K2.5 measured
APEX-AgentsNot measured47.5%Only Qwen3.8 27B measured
OSWorld63.3%Not measuredOnly Kimi K2.5 measured
Vending-Bench 2$1,198Not measuredOnly Kimi K2.5 measured
Neither model has a result on the other 21 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index29
Reasoning benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
GPQA Diamond84.1%88.9%Qwen3.8 27B
LiveBench reasoningNot measured80.0%Only Qwen3.8 27B measured
MMMU Pro (Vals AI)84.3%83.9%Kimi K2.5
GPQA Diamond (Artificial Analysis run)87.9%90.5%Qwen3.8 27B
Humanity's Last Exam (Artificial Analysis run)30.7%33.9%Qwen3.8 27B
ARC-AGI65.3%87.5%Qwen3.8 27B
ARC-AGI-211.8%42.4%Qwen3.8 27B
SimpleBench46.8%Not measuredOnly Kimi K2.5 measured
LMArena Text: hard prompts14531459Qwen3.8 27B
LMArena Text: expert14661479Qwen3.8 27B
LMArena Text: hard prompts in English14601461Qwen3.8 27B
LMArena Vision: diagrams12821285Qwen3.8 27B
LMArena Vision: homework12971295Kimi K2.5
Neither model has a result on the other 8 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index20
Math benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
AIME95.6%Not measuredOnly Kimi K2.5 measured
ProofBenchNot measured16.0%Only Qwen3.8 27B measured
LMArena Text: math14701455Kimi K2.5
LMArena Text: mathematical work14731458Kimi K2.5
Neither model has a result on the other 9 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index53
Knowledge benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
MMLU Pro85.9%84.3%Kimi K2.5
LegalBench (Vals AI)Not measured82.4%Only Qwen3.8 27B measured
CaseLaw (Vals AI)58.7%Not measuredOnly Kimi K2.5 measured
CorpFin (Vals AI)68.3%Not measuredOnly Kimi K2.5 measured
TaxEval (Vals AI)74.2%70.8%Kimi K2.5
MortgageTax (Vals AI)66.5%64.9%Kimi K2.5
MedQA (Vals AI)94.4%Not measuredOnly Kimi K2.5 measured
MedCode (Vals AI)39.3%28.7%Kimi K2.5
SimpleQA Verified34.3%Not measuredOnly Kimi K2.5 measured
LMArena Text: science14591467Qwen3.8 27B
LMArena Text: medicine and healthcare14481437Kimi K2.5
LMArena Text: legal and government14411447Qwen3.8 27B
LMArena Text: business and finance14271438Qwen3.8 27B
Neither model has a result on the other 4 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
Web design benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
Design Arena: websites1249Not measuredOnly Kimi K2.5 measured
Design Arena: UI components1236Not measuredOnly Kimi K2.5 measured
Design Arena: game dev1207Not measuredOnly Kimi K2.5 measured
Design Arena: data viz1219Not measuredOnly Kimi K2.5 measured
Design Arena: 3D1205Not measuredOnly Kimi K2.5 measured
Design Arena: all code categories1236Not measuredOnly Kimi K2.5 measured
Design Arena: SVG1149Not measuredOnly Kimi K2.5 measured
Design Arena: ASCII art1165Not measuredOnly Kimi K2.5 measured
LiveBench designNot measured75.9%Only Qwen3.8 27B measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index1610
General benchmarks for Kimi K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
LiveBench data analysisNot measured76.6%Only Qwen3.8 27B measured
SAGE (Vals AI)49.9%52.4%Qwen3.8 27B
MedScribe (Vals AI)76.4%83.8%Qwen3.8 27B
IFBench70.2%Not measuredOnly Kimi K2.5 measured
Long context reasoning (AA-LCR)78.0%82.0%Qwen3.8 27B
DTBenchNot measured88.0%Only Qwen3.8 27B measured
LMCANot measured41.4%Only Qwen3.8 27B measured
Fiction.LiveBench86.1%Not measuredOnly Kimi K2.5 measured
CL-bench19.3%Not measuredOnly Kimi K2.5 measured
Surface Evolver BenchNot measured45.0%Only Qwen3.8 27B measured
CL-bench Life13.2%Not measuredOnly Kimi K2.5 measured
LMArena Text14451440Kimi K2.5
LMArena Text: instruction following14311436Qwen3.8 27B
LMArena Text: creative writing14221383Kimi K2.5
LMArena Vision12691269Qwen3.8 27B
LMArena Text: English14531448Kimi K2.5
LMArena Text: non-English14331428Kimi K2.5
LMArena Text: Chinese14941501Qwen3.8 27B
LMArena Text: German14411443Qwen3.8 27B
LMArena Text: Japanese14231387Kimi K2.5
LMArena Text: Korean14121384Kimi K2.5
LMArena Text: French14541465Qwen3.8 27B
LMArena Text: Spanish14501449Kimi K2.5
LMArena Text: Russian14351415Kimi K2.5
LMArena Text: Polish14451439Kimi K2.5
LMArena Text: multi-turn14441440Kimi K2.5
LMArena Text: longer queries14451448Qwen3.8 27B
LMArena Text: writing and language14251405Kimi K2.5
LMArena Text: entertainment and media14171393Kimi K2.5
LMArena Vision: English12741267Kimi K2.5
LMArena Vision: Chinese13281326Kimi K2.5
LMArena Vision: reading text12761281Qwen3.8 27B
LMArena Vision: humor12841235Kimi K2.5
LMArena Vision: captioning1257Not measuredOnly Kimi K2.5 measured
LMArena Vision: recognizing things1285Not measuredOnly Kimi K2.5 measured
LMArena Vision: creative writing12691237Kimi K2.5
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 K2.5 and Qwen3.8 27B
BenchmarkKimi K2.5Qwen3.8 27BLeader
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 K2.5$0.0018Fits in one request
Qwen3.8 27B$0.0007Fits in one request

Qwen3.8 27B costs less

Repository review

50K input + 3K output tokens

Kimi K2.5$0.033Fits in one request
Qwen3.8 27B$0.0060Fits in one request

Qwen3.8 27B costs less

Agent loop

200K input + 10K output tokens

Kimi K2.5$0.125Fits in one request
Qwen3.8 27B$0.021Fits in one request

Qwen3.8 27B 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 K2.5 and Qwen3.8 27B
SpecificationMoonshotAI Kimi K2.5Qwen3.8 27B
MakerMoonshot AIQwen
WeightsOpen weightsOpen weights
ReleasedJanuary 27, 2026August 14, 2026
Context window262K1M
Input $/Mtok$0.50$0.04
Output $/Mtok$2.50$1.35
Coding index58.573.6
Independent measurements behind it22
Cost per solved task$0.24$1.26
SourceOpenRouter listingOpenRouter listing

Questions

Is Kimi K2.5 better than Qwen3.8 27B for coding?

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 58.5 for Kimi K2.5. Pick Kimi K2.5 when a solved task costs $0.24 against $1.26, or when what matters is LMArena Text: coding or MMMU Pro (Vals AI), where it leads.

Which is cheaper, Kimi K2.5 or Qwen3.8 27B?

Kimi K2.5 costs $0.50 per million input tokens and $2.50 per million output tokens; Qwen3.8 27B costs $0.04 and $1.35. Per solved SWE-bench Verified task, Kimi K2.5 costs $0.24 and Qwen3.8 27B $1.26.

Which has the bigger context window, Kimi K2.5 or Qwen3.8 27B?

Qwen3.8 27B, with 1M tokens against 262K.

How many benchmarks is the Kimi K2.5 vs Qwen3.8 27B comparison based on?

64 benchmarks have a result for both models as of October 21, 2026. Kimi K2.5's coding index rests on 2 independent benchmarks and Qwen3.8 27B's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/kimi-k2-5-vs-qwen3-8-27b" title="Kimi K2.5 vs Qwen3.8 27B, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/kimi-k2-5-vs-qwen3-8-27b">LLMBenchmark.ai, Kimi K2.5 vs Qwen3.8 27B</a></p>