Enter opens the highlighted result. Esc closes.

Kimi K3 vs Phi 4

Updated October 21, 2026. 34 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Kimi K3 Replace Phi 4
Kimi K3Moonshot AI
Coding
62.3
Overall
43.6
Output $/M
$15.00
34
Phi 4Microsoft
Coding
No data
Overall
5.9
Output $/M
$0.14
0
vs boards won of 34 shared

Decision reading

Only Kimi K3 has a coding score, 62.3 on the Vibe Coding Index; Phi 4 has none, so the two cannot be ranked on coding.

Phi 4 has the edge where output tokens cost 107.1x less.

Model AKimi K3Moonshot AI62.3/1006 independent benchmarks$0.81 per solved taskModel BPhi 4MicrosoftNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
34
Kimi K3 only
83
Phi 4 only
11
Like-for-like categories
6 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 K3 62.3 from 6 benchmarks, Phi 4 not benchmarked.
Widest category
Math, 55.6 points apart on average over 3 shared benchmarks.
Widest single gap
Long context reasoning (AA-LCR): Kimi K3 88.7% against Phi 4 0.0%, 88.7 points apart on a 0 to 100 scale.
Output price
Kimi K3 $15.00 against Phi 4 $0.14 per million output tokens, 107.1x apart.
Context window
Kimi K3 takes 1.05M tokens, Phi 4 16K.

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 K3BarsPhi 4Ties
Coding20
Intelligence20
Reasoning60
Math30
Knowledge40
General170
All shared boards340
CodingLike-for-likeSix of these boards feed the Vibe Coding Index20
Coding benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
LiveBench coding59.3%Not measuredOnly Kimi K3 measured
LMArena WebDev1655Not measuredOnly Kimi K3 measured
AA Coding Index76.2Not measuredOnly Kimi K3 measured
LiveCodeBench (Vals AI)87.2%Not measuredOnly Kimi K3 measured
SWE-bench Verified (Vals AI)93.4%Not measuredOnly Kimi K3 measured
Terminal-Bench 4 (Vals AI)17.2%Not measuredOnly Kimi K3 measured
Vibe Code Bench (Vals AI)85.0%Not measuredOnly Kimi K3 measured
Code migration (Vals AI)16.1%Not measuredOnly Kimi K3 measured
ProgramBench (Vals AI)2.0%Not measuredOnly Kimi K3 measured
IOI (Vals AI)48.9%Not measuredOnly Kimi K3 measured
Terminal-Bench 2.1 (Vals AI)80.9%Not measuredOnly Kimi K3 measured
SWE-bench Multimodal (Vals AI)44.2%Not measuredOnly Kimi K3 measured
Vibe Code Bench 1-100 (Vals AI)18.2%Not measuredOnly Kimi K3 measured
CyberBench (Vals AI)75.2%Not measuredOnly Kimi K3 measured
LiveCodeBench (Artificial Analysis run)Not measured23.1%Only Phi 4 measured
SciCode59.5%Not measuredOnly Kimi K3 measured
Terminal-Bench Hard (Artificial Analysis run)Not measured3.8%Only Phi 4 measured
FrontierCode44.2%Not measuredOnly Kimi K3 measured
FrontierSWE25.9%Not measuredOnly Kimi K3 measured
WeirdML82.6%Not measuredOnly Kimi K3 measured
ALE-Bench1524.5Not measuredOnly Kimi K3 measured
LMArena Text: coding15081231Kimi K3
LMArena Text: software and IT14991228Kimi K3
LMArena WebDev: React1670Not measuredOnly Kimi K3 measured
LMArena WebDev: HTML1644Not measuredOnly Kimi K3 measured
LMArena WebDev: from an image1579Not measuredOnly Kimi K3 measured
Neither model has a result on the other 12 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
AA Intelligence Index43.65.9Kimi K3
Epoch Capabilities Index157.9130.5Kimi K3
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index26 boards, none shared
Agentic work benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
AA Agentic Index50.0Not measuredOnly Kimi K3 measured
Design Arena: agents building web apps1295Not measuredOnly Kimi K3 measured
Design Arena: agents building full-stack apps1315Not measuredOnly Kimi K3 measured
Design Arena: agents building mobile apps1268Not measuredOnly Kimi K3 measured
Design Arena: agents building Android apps1256Not measuredOnly Kimi K3 measured
Design Arena: agents building Godot games1223Not measuredOnly Kimi K3 measured
Design Arena: agents building games1252Not measuredOnly Kimi K3 measured
Design Arena: HTML slides1256Not measuredOnly Kimi K3 measured
Design Arena: PowerPoint slides1273Not measuredOnly Kimi K3 measured
Design Arena: spreadsheets1054Not measuredOnly Kimi K3 measured
LiveBench research79.0%Not measuredOnly Kimi K3 measured
LiveBench browser16.0%Not measuredOnly Kimi K3 measured
LiveBench long agentic42.8%Not measuredOnly Kimi K3 measured
Finance Agent (Vals AI)53.1%Not measuredOnly Kimi K3 measured
Tax Agent Bench (Vals AI)33.4%Not measuredOnly Kimi K3 measured
Legal Research Bench (Vals AI)46.1%Not measuredOnly Kimi K3 measured
Harvey's Legal Agent Benchmark (Vals AI)12.9%Not measuredOnly Kimi K3 measured
Excel Modeling Benchmark (Vals AI)66.7%Not measuredOnly Kimi K3 measured
BioMysteryBench (Vals AI)72.6%Not measuredOnly Kimi K3 measured
Terminal-Bench-Science (Vals AI)1.4%Not measuredOnly Kimi K3 measured
Time Horizon Index (Vals AI)10.5%Not measuredOnly Kimi K3 measured
τ²-Bench TelecomNot measured0.0%Only Phi 4 measured
APEX-Agents47.8%Not measuredOnly Kimi K3 measured
BALROGNot measured11.6%Only Phi 4 measured
PostTrainBench32.0%Not measuredOnly Kimi K3 measured
Vending-Bench 2$5,165Not measuredOnly Kimi K3 measured
Neither model has a result on the other 13 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
Reasoning benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
GPQA Diamond92.9%Not measuredOnly Kimi K3 measured
LiveBench reasoning90.7%Not measuredOnly Kimi K3 measured
MMMU Pro (Vals AI)88.2%Not measuredOnly Kimi K3 measured
GPQA Diamond (Artificial Analysis run)93.5%57.5%Kimi K3
Humanity's Last Exam (Artificial Analysis run)46.9%3.8%Kimi K3
GPQA Diamond (Epoch AI run)93.1%56.1%Kimi K3
ARC-AGI94.5%Not measuredOnly Kimi K3 measured
ARC-AGI-260.4%Not measuredOnly Kimi K3 measured
SimpleBench60.7%Not measuredOnly Kimi K3 measured
Chess puzzles37.0%Not measuredOnly Kimi K3 measured
Mystery game puzzles26.0%Not measuredOnly Kimi K3 measured
LMArena Text: hard prompts14951220Kimi K3
LMArena Text: expert15201203Kimi K3
LMArena Text: hard prompts in English14961230Kimi K3
Neither model has a result on the other 7 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index30
Math benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
AA Math IndexNot measured18.0Only Phi 4 measured
AIME 2025 (Artificial Analysis run)Not measured18.0%Only Phi 4 measured
OTIS Mock AIME97.2%13.8%Kimi K3
FrontierMath, tiers 1 to 372.2%Not measuredOnly Kimi K3 measured
FrontierMath, tier 439.0%Not measuredOnly Kimi K3 measured
ProofBench87.0%Not measuredOnly Kimi K3 measured
MATH level 5Not measured64.9%Only Phi 4 measured
LMArena Text: math14941246Kimi K3
LMArena Text: mathematical work15051254Kimi K3
Neither model has a result on the other 4 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
Knowledge benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
MMLU Pro88.0%Not measuredOnly Kimi K3 measured
LegalBench (Vals AI)86.2%Not measuredOnly Kimi K3 measured
CorpFin (Vals AI)71.6%Not measuredOnly Kimi K3 measured
TaxEval (Vals AI)75.7%Not measuredOnly Kimi K3 measured
MortgageTax (Vals AI)66.3%Not measuredOnly Kimi K3 measured
MedCode (Vals AI)49.4%Not measuredOnly Kimi K3 measured
MMLU Pro (Artificial Analysis run)Not measured71.4%Only Phi 4 measured
SimpleQA Verified50.6%Not measuredOnly Kimi K3 measured
MMLUNot measured84.8%Only Phi 4 measured
LMArena Text: science15081206Kimi K3
LMArena Text: medicine and healthcare14891202Kimi K3
LMArena Text: legal and government15031241Kimi K3
LMArena Text: business and finance14701199Kimi K3
Neither model has a result on the other 4 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Web design benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
Design Arena: websites1335Not measuredOnly Kimi K3 measured
Design Arena: UI components1344Not measuredOnly Kimi K3 measured
Design Arena: game dev1360Not measuredOnly Kimi K3 measured
Design Arena: data viz1350Not measuredOnly Kimi K3 measured
Design Arena: 3D1388Not measuredOnly Kimi K3 measured
Design Arena: all code categories1361Not measuredOnly Kimi K3 measured
Design Arena: SVG1302Not measuredOnly Kimi K3 measured
LiveBench design77.2%Not measuredOnly Kimi K3 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index170
General benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
Vals Index53.0%Not measuredOnly Kimi K3 measured
LiveBench data analysis78.7%Not measuredOnly Kimi K3 measured
Public Benefits Bench (Vals AI)68.2%Not measuredOnly Kimi K3 measured
SAGE (Vals AI)52.8%Not measuredOnly Kimi K3 measured
MedScribe (Vals AI)88.0%Not measuredOnly Kimi K3 measured
Vals Multimodal Index73.4%Not measuredOnly Kimi K3 measured
IFBenchNot measured23.5%Only Phi 4 measured
Long context reasoning (AA-LCR)88.7%0.0%Kimi K3
DTBench91.2%Not measuredOnly Kimi K3 measured
LMCA52.7%Not measuredOnly Kimi K3 measured
Creative writing (Lech Mazur)Not measured62.6%Only Phi 4 measured
Surface Evolver Bench95.0%Not measuredOnly Kimi K3 measured
Furniture assembly34.2%Not measuredOnly Kimi K3 measured
LMArena Text14761217Kimi K3
LMArena Text: instruction following14851201Kimi K3
LMArena Text: creative writing14571181Kimi K3
LMArena Text: English14811231Kimi K3
LMArena Text: non-English14661197Kimi K3
LMArena Text: Chinese15291212Kimi K3
LMArena Text: German14901222Kimi K3
LMArena Text: Japanese14941158Kimi K3
LMArena Text: Korean14601151Kimi K3
LMArena Text: French14921224Kimi K3
LMArena Text: Spanish14721234Kimi K3
LMArena Text: Russian14851209Kimi K3
LMArena Text: Polish1458Not measuredOnly Kimi K3 measured
LMArena Text: multi-turn14891206Kimi K3
LMArena Text: longer queries14941217Kimi K3
LMArena Text: writing and language14711198Kimi K3
LMArena Text: entertainment and media14531162Kimi K3
Neither model has a result on the other 16 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Kimi K3 and Phi 4
BenchmarkKimi K3Phi 4Leader
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 K3$0.011Fits in one request
Phi 4$0.0001Fits in one request

Phi 4 costs less

Repository review

50K input + 3K output tokens

Kimi K3$0.195Fits in one request
Phi 4$0.0039Exceeds the context window

Phi 4 costs less

Agent loop

200K input + 10K output tokens

Kimi K3$0.750Fits in one request
Phi 4$0.015Exceeds the context window

Phi 4 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 K3 and Phi 4
SpecificationMoonshotAI Kimi K3Microsoft Phi 4
MakerMoonshot AIMicrosoft
WeightsOpen weightsOpen weights
ReleasedJuly 16, 2026January 10, 2025
Context window1.05M16K
Input $/Mtok$3.00$0.07
Output $/Mtok$15.00$0.14
Coding index62.3Not benchmarked
Independent measurements behind it60
Cost per solved task$0.81Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Kimi K3 better than Phi 4 for coding?

Only Kimi K3 has a coding score, 62.3 on the Vibe Coding Index; Phi 4 has none, so the two cannot be ranked on coding. Phi 4 has the edge where output tokens cost 107.1x less.

Which is cheaper, Kimi K3 or Phi 4?

Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens; Phi 4 costs $0.07 and $0.14. Cost per solved task is known for Kimi K3 only.

Which has the bigger context window, Kimi K3 or Phi 4?

Kimi K3, with 1.05M tokens against 16K.

How many benchmarks is the Kimi K3 vs Phi 4 comparison based on?

34 benchmarks have a result for both models as of October 21, 2026. Kimi K3's coding index rests on 6 independent benchmarks and Phi 4's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next