Skip to main content

Model comparison

Kimi K2.5 vs Qwen3.7 Plus

Data verified

Head-to-head evidence from 38 shared benchmark results across 8 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

Moonshot AI
59.66/100
Margin
7.6pts
winning →
67.22/100
1 category wins7 category wins

Public leaderboard positions: Kimi K2.5 #54 (Supported); Qwen3.7 Plus #21 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Kimi K2.5 and Qwen3.7 Plus share 38 comparable benchmark results. 8 of 8 categories are comparable. 25 results are unique to Kimi K2.5; 31 to Qwen3.7 Plus.

Updated July 23, 2026
Shared results
38
Kimi K2.5 only
25
Qwen3.7 Plus only
31
Comparable categories
8 / 8

Pick Qwen3.7 Plus if you want the stronger benchmark profile. Kimi K2.5 only becomes the better choice if instruction following is the priority or you would rather avoid the extra latency and token burn of a reasoning model.

Why this result

Qwen3.7 Plus is clearly ahead on the BenchAlign aggregate, 67.22 to 59.66. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Qwen3.7 Plus's sharpest advantage is in mathematics, where it averages 92.9 against 60.6. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 50.8% to 70.3%. Kimi K2.5 does hit back in instruction following, so the answer changes if that is the part of the workload you care about most.

Qwen3.7 Plus is the reasoning model in the pair, while Kimi K2.5 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Qwen3.7 Plus gives you the larger context window at 1M, compared with 256K for Kimi K2.5.

Category breakdown

Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.

Category scores and score margins for Kimi K2.5 and Qwen3.7 Plus
CategoryKimi K2.5ΔQwen3.7 Plus
MathKimi K2.560.6Margin 32.3Qwen3.7 Plus92.9
ReasoningKimi K2.561.0Margin 30.7Qwen3.7 Plus91.7
AgenticKimi K2.555.0Margin 16.7Qwen3.7 Plus71.7
CodingKimi K2.559.4Margin 16.2Qwen3.7 Plus75.6
Inst. FollowingKimi K2.593.9Margin 9.4Qwen3.7 Plus84.5
KnowledgeKimi K2.556.9Margin 3.2Qwen3.7 Plus60.1
MultilingualKimi K2.582.3Margin 3.1Qwen3.7 Plus85.4
MultimodalKimi K2.578.5Margin 3.0Qwen3.7 Plus81.5

Decisive benchmark drivers

The largest measured benchmark gaps in this matchup, with exact reported values.

More
A · Kimi K2.5B · Qwen3.7 Plus
  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 50.8%B 70.3%
    Winner: Qwen3.7 PlusΔ 19.5
    Terminal-Bench 2.0: Kimi K2.5 scored 50.8%; Qwen3.7 Plus scored 70.3%. Qwen3.7 Plus wins this benchmark.
  2. SWE-bench Pro

    Coding
    Source ↗
    A 50.7%B 57.6%
    Winner: Qwen3.7 PlusΔ 6.9
    SWE-bench Pro: Kimi K2.5 scored 50.7%; Qwen3.7 Plus scored 57.6%. Qwen3.7 Plus wins this benchmark.
  3. HMMT Feb 2026

    Math
    Source ↗
    A 87.1%B 92.9%
    Winner: Qwen3.7 PlusΔ 5.8
    HMMT Feb 2026: Kimi K2.5 scored 87.1%; Qwen3.7 Plus scored 92.9%. Qwen3.7 Plus wins this benchmark.
  4. HLE

    Knowledge
    Source ↗
    A 30.1%B 34.7%
    Winner: Qwen3.7 PlusΔ 4.6
    HLE: Kimi K2.5 scored 30.1%; Qwen3.7 Plus scored 34.7%. Qwen3.7 Plus wins this benchmark.
  5. MMLU-ProX

    Multilingual
    Source ↗
    A 82.3%B 85.4%
    Winner: Qwen3.7 PlusΔ 3.1
    MMLU-ProX: Kimi K2.5 scored 82.3%; Qwen3.7 Plus scored 85.4%. Qwen3.7 Plus wins this benchmark.

Operational comparison

Runtime and commercial metrics are compared only when both models have a complete sourced value.

MetricKimi K2.5Qwen3.7 PlusComparison
Input / output priceUSD per 1M tokensKimi K2.5$0.6 input / $3 outputQwen3.7 PlusNot availableA complete price comparison is not available.
Generation speedtokens per secondKimi K2.545 tok/sQwen3.7 PlusNot availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenKimi K2.52.38 sQwen3.7 PlusNot availableA complete latency comparison is not available.
Context windowmaximum listed tokensKimi K2.5256KQwen3.7 Plus1MQwen3.7 Plus lists the larger context window.

Benchmark Deep Dive

AgenticQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
Terminal-Bench 2.0Source 50.8%70.3%Qwen3.7 Plus leads
BrowseCompSource 60.6%Not comparable
Claw-EvalSource 52.3%62.7%Qwen3.7 Plus leads
QwenClawBenchSource 54.3%61.8%Qwen3.7 Plus leads
τ³-bench resultsSource 65.7%Not comparable
DeepSearchQASource 77.1%Not comparable
DeepPlanningSource 14.4%62.3%Qwen3.7 Plus leads
ToolathlonSource 27.8%Not comparable
MCP AtlasSource 29.5%73.2%Qwen3.7 Plus leads
MCP-TasksSource 59.1%Not comparable
WideResearchSource 72.7%Not comparable
τ²-bench resultsSource 95.9%93%Kimi K2.5 leads
APEX-Agents-AASource 11.5%22.4%Qwen3.7 Plus leads
Gert LabsSource 45.88%Not comparable
ResearchClawBenchSource 14.0%Not comparable
JobBenchSource 8.7%Not comparable
AA Agentic IndexSource 21.7%20.8%Kimi K2.5 leads
GDPval-AASource 25.4%21.8%Kimi K2.5 leads
GDPval-AASource 1009936Kimi K2.5 leads
QwenWebBenchSource 1536Not comparable
BFCL v4Source 72.9%Not comparable
VITA-BenchSource 45.6%Not comparable
OSWorld-VerifiedSource 73.3%Not comparable
AndroidWorldSource 81.0%Not comparable
OSWorld 2.0Source 2.8%Not comparable
CodingQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
SWE-bench VerifiedSource 76.8%77.7%Qwen3.7 Plus leads
SWE-bench Verified*Source 70.8%Not comparable
LiveCodeBench v6Source 85.0%Not comparable
SWE-bench ProSource 50.7%57.6%Qwen3.7 Plus leads
SWE MultilingualSource 73%75.8%Qwen3.7 Plus leads
SWE-RebenchSource 58.5%Not comparable
React Native EvalsSource 77.2%Not comparable
SciCodeSource 48.7%51.3%Qwen3.7 Plus leads
AA-SciCodeSource 49.0%45.5%Kimi K2.5 leads
AA Coding IndexSource 46.8%55.9%Qwen3.7 Plus leads
Terminal-Bench 2.0Source 70.3%Not comparable
NL2RepoSource 41.1%Not comparable
LiveCodeBenchSource 89.6%Not comparable
ReasoningQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
LongBench v2Source 61%Not comparable
AA-LCRSource 65.3%65.0%Kimi K2.5 leads
CritPtSource 3.1%9.1%Qwen3.7 Plus leads
MRCRv2Source 91.7%Not comparable
KnowledgeQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
GPQASource 87.6%90.3%Qwen3.7 Plus leads
GPQA-DSource 87.6%90.3%Qwen3.7 Plus leads
SuperGPQASource 69.2%71.4%Qwen3.7 Plus leads
MMLU-ProSource 87.1%88.5%Qwen3.7 Plus leads
MMLU-Pro (Arcee)Source 87.1%Not comparable
HLESource 30.1%34.7%Qwen3.7 Plus leads
Artificial Analysis Intelligence IndexSource 35.4%39.0%Qwen3.7 Plus leads
AA-GPQA DiamondSource 87.9%90.0%Qwen3.7 Plus leads
AA-HLESource 29.4%33.4%Qwen3.7 Plus leads
AA-Omniscience IndexSource -8.1%2.4%Qwen3.7 Plus leads
AA-Omniscience AccuracySource 34.3%22.2%Kimi K2.5 leads
AA-Omniscience Hallucination RateSource 64.6%25.5%Qwen3.7 Plus leads
MMLU-ReduxSource 94.5%Not comparable
MMMLUSource 89.0%Not comparable
MathQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
AIME 2025Source 96.1%Not comparable
AIME26Source 95.8%Not comparable
AIME25 (Arcee)Source 96.3%Not comparable
HMMT Feb 2025Source 95.4%Not comparable
HMMT Nov 2025Source 91.1%Not comparable
HMMT Feb 2026Source 87.1%92.9%Qwen3.7 Plus leads
MMAnswerBenchSource 81.8%Not comparable
FrontierMath v2 (Tiers 1-3)Source 27.900%Not comparable
FrontierMath v2 (Tier 4)Source 4.200%Not comparable
IMOAnswerBenchSource 86.0%Not comparable
ApexSource 22.7%Not comparable
MultilingualQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
MMLU-ProXSource 82.3%85.4%Qwen3.7 Plus leads
NOVA-63Source 56.0%58.8%Qwen3.7 Plus leads
INCLUDESource 83.0%Not comparable
MAXIFESource 88.8%Not comparable
PolyMathSource 84.0%Not comparable
MultimodalQwen3.7 Plus wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
MMMU-ProSource 78.5%79%Qwen3.7 Plus leads
Video-MMESource 87.4%Not comparable
MMVUSource 80.4%Not comparable
VideoMMMUSource 86.6%85.4%Kimi K2.5 leads
AA-MMMU-ProSource 75.4%80.5%Qwen3.7 Plus leads
Design Arena WebsiteSource 12791288Qwen3.7 Plus leads
MathVisionSource 90.3%Not comparable
CharXivSource 85.9%Not comparable
ERQASource 69.8%Not comparable
MedXpertQA (MM)Source 71.0%Not comparable
ScreenSpot ProSource 79.0%Not comparable
SimpleVQASource 81.7%Not comparable
MMSearch-PlusSource 41.4%Not comparable
RealWorldQASource 86.9%Not comparable
OmniDocBench 1.5Source 91.4%Not comparable
OCRBench V2Source 70.7%Not comparable
ODINW13Source 51.1%Not comparable
Video-MME (with subtitle)Source 88.0%Not comparable
MLVU (M-Avg)Source 87.4%Not comparable
Inst. FollowingKimi K2.5 wins
BenchmarkKimi K2.5Qwen3.7 PlusResult
IFEvalSource 93.9%94.6%Qwen3.7 Plus leads
AA-IFBenchSource 70.2%78.0%Qwen3.7 Plus leads
IFBenchSource 79.1%Not comparable
Frequently Asked Questions (9)

Which is better, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus is ahead on BenchLM's BenchAlign leaderboard, 67.22 to 59.66. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 50.8% and 70.3%.

Which is better for knowledge tasks, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for knowledge tasks in this comparison, averaging 60.1 versus 56.9. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.

Which is better for coding, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for coding in this comparison, averaging 75.6 versus 59.4. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.

Which is better for math, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for math in this comparison, averaging 92.9 versus 60.6. Inside this category, HMMT Feb 2026 is the benchmark that creates the most daylight between them.

Which is better for reasoning, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for reasoning in this comparison, averaging 91.7 versus 61. Inside this category, CritPt is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for agentic tasks in this comparison, averaging 71.7 versus 55. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.

Which is better for multimodal and grounded tasks, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for multimodal and grounded tasks in this comparison, averaging 81.5 versus 78.5. Inside this category, Design Arena Website is the benchmark that creates the most daylight between them.

Which is better for instruction following, Kimi K2.5 or Qwen3.7 Plus?

Kimi K2.5 has the edge for instruction following in this comparison, averaging 93.9 versus 84.5. Inside this category, AA-IFBench is the benchmark that creates the most daylight between them.

Which is better for multilingual tasks, Kimi K2.5 or Qwen3.7 Plus?

Qwen3.7 Plus has the edge for multilingual tasks in this comparison, averaging 85.4 versus 82.3. Inside this category, MMLU-ProX is the benchmark that creates the most daylight between them.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

Kimi K2.5
API / mo$2,700
Self-host / mo$5,221
Break-even132M/day
Qwen3.7 Plus
API / mo$0
Self-host / moNot listed
Break-even
Proprietary model — self-hosting not applicable.
Model the full break-even

Related Comparisons

Last updated: July 23, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.