Skip to main content
BenchLM
Advertise

Model comparison

Kimi K2.5 vs Qwen3.7 Max

Data verified

Head-to-head evidence from 35 shared benchmark results across 8 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

Kimi K2.5
Moonshot AI
59.66/100
Margin
13.2pts
winning →
72.84/100
1 category wins6 category wins

Public leaderboard positions: Kimi K2.5 #54 (Supported); Qwen3.7 Max #10 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Kimi K2.5 and Qwen3.7 Max share 35 comparable benchmark results. 7 of 8 categories are comparable. 28 results are unique to Kimi K2.5; 23 to Qwen3.7 Max.

Updated July 23, 2026
Shared results
35
Kimi K2.5 only
28
Qwen3.7 Max only
23
Comparable categories
7 / 8

Pick Qwen3.7 Max 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.

Confidence note. This is a partial-evidence comparison with 35 shared benchmark results across 8 evidence categories; 7 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Qwen3.7 Max is clearly ahead on the BenchAlign aggregate, 72.84 to 59.66. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Qwen3.7 Max's sharpest advantage is in mathematics, where it averages 97.1 against 60.6. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 50.8% to 69.7%. 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 Max 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 Max 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 Max
CategoryKimi K2.5ΔQwen3.7 Max
MathKimi K2.560.6Margin→ 36.5Qwen3.7 Max97.1
ReasoningKimi K2.561.0Margin→ 29.4Qwen3.7 Max90.4
CodingKimi K2.559.4Margin→ 18.5Qwen3.7 Max77.9
AgenticKimi K2.555.0Margin→ 14.7Qwen3.7 Max69.7
Inst. FollowingKimi K2.593.9Margin← 9.5Qwen3.7 Max84.4
KnowledgeKimi K2.556.9Margin→ 7.3Qwen3.7 Max64.2
MultilingualKimi K2.582.3Margin→ 4.7Qwen3.7 Max87.0
MultimodalKimi K2.578.5MarginNo overlapQwen3.7 MaxNot measured

Decisive benchmark drivers

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

More
A · Kimi K2.5B · Qwen3.7 Max

  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 50.8%B 69.7%
    Winner: Qwen3.7 MaxΔ 18.9
    Terminal-Bench 2.0: Kimi K2.5 scored 50.8%; Qwen3.7 Max scored 69.7%. Qwen3.7 Max wins this benchmark.

  2. HLE

    Knowledge
    Source ↗
    A 30.1%B 41.4%
    Winner: Qwen3.7 MaxΔ 11.3
    HLE: Kimi K2.5 scored 30.1%; Qwen3.7 Max scored 41.4%. Qwen3.7 Max wins this benchmark.

  3. HMMT Feb 2026

    Math
    Source ↗
    A 87.1%B 97.1%
    Winner: Qwen3.7 MaxΔ 10
    HMMT Feb 2026: Kimi K2.5 scored 87.1%; Qwen3.7 Max scored 97.1%. Qwen3.7 Max wins this benchmark.

  4. SWE-bench Pro

    Coding
    Source ↗
    A 50.7%B 60.6%
    Winner: Qwen3.7 MaxΔ 9.9
    SWE-bench Pro: Kimi K2.5 scored 50.7%; Qwen3.7 Max scored 60.6%. Qwen3.7 Max wins this benchmark.

  5. GPQA

    Knowledge
    Source ↗
    A 87.6%B 92.4%
    Winner: Qwen3.7 MaxΔ 4.8
    GPQA: Kimi K2.5 scored 87.6%; Qwen3.7 Max scored 92.4%. Qwen3.7 Max wins this benchmark.

Operational comparison

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

MetricKimi K2.5Qwen3.7 MaxComparison
Input / output priceUSD per 1M tokensKimi K2.5$0.6 input / $3 outputQwen3.7 MaxNot availableA complete price comparison is not available.
Generation speedtokens per secondKimi K2.545 tok/sQwen3.7 MaxNot availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenKimi K2.52.38 sQwen3.7 MaxNot availableA complete latency comparison is not available.
Context windowmaximum listed tokensKimi K2.5256KQwen3.7 Max1MQwen3.7 Max lists the larger context window.

Benchmark Deep Dive

AgenticQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
Terminal-Bench 2.0Source 50.8%69.7%Qwen3.7 Max leads
BrowseCompSource 60.6%—Not comparable
Claw-EvalSource 52.3%65.2%Qwen3.7 Max leads
QwenClawBenchSource 54.3%64.3%Qwen3.7 Max leads
τ³-bench resultsSource 65.7%—Not comparable
DeepSearchQASource 77.1%—Not comparable
DeepPlanningSource 14.4%—Not comparable
ToolathlonSource 27.8%—Not comparable
MCP AtlasSource 29.5%76.4%Qwen3.7 Max leads
MCP-TasksSource 59.1%—Not comparable
WideResearchSource 72.7%—Not comparable
τ²-bench resultsSource 95.9%94.7%Kimi K2.5 leads
APEX-Agents-AASource 11.5%—Not comparable
Gert LabsSource 45.88%64.27%Qwen3.7 Max leads
ResearchClawBenchSource 14.0%18.7%Qwen3.7 Max leads
JobBenchSource 8.7%—Not comparable
AA Agentic IndexSource 21.7%30.6%Qwen3.7 Max leads
GDPval-AASource 25.4%38.7%Qwen3.7 Max leads
GDPval-AASource 10091273Qwen3.7 Max leads
QwenWebBenchSource —1568Not comparable
BFCL v4Source —75.0%Not comparable
VITA-BenchSource —47.9%Not comparable
HLE w/ toolsSource —53.5%Not comparable
AA BriefcaseSource —908Not comparable
AA AutomationBenchSource —25.6%Not comparable
AA EnterpriseOps-GymSource —45.0%Not comparable
AA ITBenchSource —42.5%Not comparable
terminalBenchHardSource —50.8%Not comparable
aaTerminalBench21Source —74.5%Not comparable
AA Harvey LABSource —83.4%Not comparable
CodingQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
SWE-bench VerifiedSource 76.8%80.4%Qwen3.7 Max leads
SWE-bench Verified*Source 70.8%—Not comparable
LiveCodeBench v6Source 85.0%—Not comparable
SWE-bench ProSource 50.7%60.6%Qwen3.7 Max leads
SWE MultilingualSource 73%78.3%Qwen3.7 Max leads
SWE-RebenchSource 58.5%—Not comparable
React Native EvalsSource 77.2%—Not comparable
SciCodeSource 48.7%53.5%Qwen3.7 Max leads
AA-SciCodeSource 49.0%48.8%Kimi K2.5 leads
AA Coding IndexSource 46.8%66.0%Qwen3.7 Max leads
NL2RepoSource —47.2%Not comparable
LiveCodeBenchSource —91.6%Not comparable
Terminal-Bench 2.0Source —69.7%Not comparable
ReasoningQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
LongBench v2Source 61%—Not comparable
AA-LCRSource 65.3%69.0%Qwen3.7 Max leads
CritPtSource 3.1%13.4%Qwen3.7 Max leads
MRCRv2Source —90.4%Not comparable
KnowledgeQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
GPQASource 87.6%92.4%Qwen3.7 Max leads
GPQA-DSource 87.6%92.4%Qwen3.7 Max leads
SuperGPQASource 69.2%73.6%Qwen3.7 Max leads
MMLU-ProSource 87.1%89.6%Qwen3.7 Max leads
MMLU-Pro (Arcee)Source 87.1%—Not comparable
HLESource 30.1%41.4%Qwen3.7 Max leads
Artificial Analysis Intelligence IndexSource 35.4%46.0%Qwen3.7 Max leads
AA-GPQA DiamondSource 87.9%92.3%Qwen3.7 Max leads
AA-HLESource 29.4%38.1%Qwen3.7 Max leads
AA-Omniscience IndexSource -8.1%14.1%Qwen3.7 Max leads
AA-Omniscience AccuracySource 34.3%30.1%Kimi K2.5 leads
AA-Omniscience Hallucination RateSource 64.6%22.9%Qwen3.7 Max leads
MMLU-ReduxSource —95%Not comparable
MMMLUSource —90.3%Not comparable
MathQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
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%97.1%Qwen3.7 Max 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 —90.0%Not comparable
ApexSource —44.5%Not comparable
MultilingualQwen3.7 Max wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
MMLU-ProXSource 82.3%87%Qwen3.7 Max leads
NOVA-63Source 56.0%59.0%Qwen3.7 Max leads
INCLUDESource —86.2%Not comparable
MAXIFESource —89.2%Not comparable
PolyMathSource —86.5%Not comparable
Multimodal
BenchmarkKimi K2.5Qwen3.7 MaxResult
MMMU-ProSource 78.5%—Not comparable
Video-MMESource 87.4%—Not comparable
MMVUSource 80.4%—Not comparable
VideoMMMUSource 86.6%—Not comparable
AA-MMMU-ProSource 75.4%—Not comparable
Design Arena WebsiteSource 12791293Qwen3.7 Max leads
Inst. FollowingKimi K2.5 wins
BenchmarkKimi K2.5Qwen3.7 MaxResult
IFEvalSource 93.9%94.3%Qwen3.7 Max leads
AA-IFBenchSource 70.2%80.5%Qwen3.7 Max leads
IFBenchSource —79.1%Not comparable
Frequently Asked Questions (8)

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

Qwen3.7 Max is ahead on BenchLM's BenchAlign leaderboard, 72.84 to 59.66. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 50.8% and 69.7%.

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

Qwen3.7 Max has the edge for knowledge tasks in this comparison, averaging 64.2 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 Max?

Qwen3.7 Max has the edge for coding in this comparison, averaging 77.9 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 Max?

Qwen3.7 Max has the edge for math in this comparison, averaging 97.1 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 Max?

Qwen3.7 Max has the edge for reasoning in this comparison, averaging 90.4 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 Max?

Qwen3.7 Max has the edge for agentic tasks in this comparison, averaging 69.7 versus 55. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.

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

Kimi K2.5 has the edge for instruction following in this comparison, averaging 93.9 versus 84.4. 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 Max?

Qwen3.7 Max has the edge for multilingual tasks in this comparison, averaging 87 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 Max
API / mo$0
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Model the full break-even

Related Comparisons

Explore More

Moonshot AI Compare Pricing Methodology Find Your Best LLM Overall Rankings
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.