Skip to main content
BenchLM
Advertise

Model comparison

Kimi K2.5 vs Qwen3.5 397B

Data verified

Head-to-head evidence from 47 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
2.6pts
← winning
57.01/100
2 category wins6 category wins

Public leaderboard positions: Kimi K2.5 #54 (Supported); Qwen3.5 397B #71 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Kimi K2.5 and Qwen3.5 397B share 47 comparable benchmark results. 8 of 8 categories are comparable. 16 results are unique to Kimi K2.5; 8 to Qwen3.5 397B.

Updated July 20, 2026
Shared results
47
Kimi K2.5 only
16
Qwen3.5 397B only
8
Comparable categories
8 / 8

Pick Kimi K2.5 if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if mathematics is the priority.

Why this result

Kimi K2.5 has the cleaner BenchAlign overall profile here, landing at 59.66 versus 57.01. It is a real lead, but still close enough that category-level strengths matter more than the headline number.

Kimi K2.5's sharpest advantage is in instruction following, where it averages 93.9 against 92.6. The single biggest benchmark swing on the page is AIME26, 95.8% to 93.3%. Qwen3.5 397B does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.

Qwen3.5 397B is also the more expensive model on tokens at $0.60 input / $3.60 output per 1M tokens, versus $0.60 input / $3.00 output per 1M tokens for Kimi K2.5. Kimi K2.5 gives you the larger context window at 256K, compared with 128K for Qwen3.5 397B.

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 Qwen3.5 397B
CategoryKimi K2.5ΔQwen3.5 397B
MathKimi K2.560.6Margin→ 30.0Qwen3.5 397B90.6
CodingKimi K2.559.4Margin→ 7.1Qwen3.5 397B66.5
MultilingualKimi K2.582.3Margin→ 2.4Qwen3.5 397B84.7
ReasoningKimi K2.561.0Margin→ 2.2Qwen3.5 397B63.2
AgenticKimi K2.555.0Margin→ 1.5Qwen3.5 397B56.5
Inst. FollowingKimi K2.593.9Margin← 1.3Qwen3.5 397B92.6
MultimodalKimi K2.578.5Margin→ 1.1Qwen3.5 397B79.6
KnowledgeKimi K2.556.9Margin← 0.3Qwen3.5 397B56.6

Decisive benchmark drivers

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

More
A · Kimi K2.5B · Qwen3.5 397B

  1. AIME26

    Math
    Source ↗
    A 95.8%B 93.3%
    Winner: Kimi K2.5Δ 2.5
    AIME26: Kimi K2.5 scored 95.8%; Qwen3.5 397B scored 93.3%. Kimi K2.5 wins this benchmark.

  2. MMLU-ProX

    Multilingual
    Source ↗
    A 82.3%B 84.7%
    Winner: Qwen3.5 397BΔ 2.4
    MMLU-ProX: Kimi K2.5 scored 82.3%; Qwen3.5 397B scored 84.7%. Qwen3.5 397B wins this benchmark.

  3. LongBench v2

    Reasoning
    Source ↗
    A 61%B 63.2%
    Winner: Qwen3.5 397BΔ 2.2
    LongBench v2: Kimi K2.5 scored 61%; Qwen3.5 397B scored 63.2%. Qwen3.5 397B wins this benchmark.

  4. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 50.8%B 52.5%
    Winner: Qwen3.5 397BΔ 1.7
    Terminal-Bench 2.0: Kimi K2.5 scored 50.8%; Qwen3.5 397B scored 52.5%. Qwen3.5 397B wins this benchmark.

  5. HLE

    Knowledge
    Source ↗
    A 30.1%B 28.7%
    Winner: Kimi K2.5Δ 1.4
    HLE: Kimi K2.5 scored 30.1%; Qwen3.5 397B scored 28.7%. Kimi K2.5 wins this benchmark.

Operational comparison

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

MetricKimi K2.5Qwen3.5 397BComparison
Input / output priceUSD per 1M tokensKimi K2.5$0.6 input / $3 outputQwen3.5 397B$0.6 input / $3.6 outputKimi K2.5 has the lower combined listed price.
Generation speedtokens per secondKimi K2.545 tok/sQwen3.5 397B96 tok/sQwen3.5 397B has the higher measured throughput.
First-answer latencyseconds to first tokenKimi K2.52.38 sQwen3.5 397B2.44 sKimi K2.5 reaches the first token sooner.
Context windowmaximum listed tokensKimi K2.5256KQwen3.5 397B128KKimi K2.5 lists the larger context window.

Benchmark Deep Dive

AgenticQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
Terminal-Bench 2.0Source 50.8%52.5%Qwen3.5 397B leads
BrowseCompSource 60.6%62%Qwen3.5 397B leads
Claw-EvalSource 52.3%56.8%Qwen3.5 397B leads
QwenClawBenchSource 54.3%51.8%Kimi K2.5 leads
τ³-bench resultsSource 65.7%68.4%Qwen3.5 397B leads
DeepSearchQASource 77.1%—Not comparable
DeepPlanningSource 14.4%37.6%Qwen3.5 397B leads
ToolathlonSource 27.8%36.3%Qwen3.5 397B leads
MCP AtlasSource 29.5%46.1%Qwen3.5 397B leads
MCP-TasksSource 59.1%74.2%Qwen3.5 397B leads
WideResearchSource 72.7%74.0%Qwen3.5 397B leads
τ²-bench resultsSource 95.9%95.6%Kimi K2.5 leads
APEX-Agents-AASource 11.5%15.3%Qwen3.5 397B leads
Gert LabsSource 45.88%46.76%Qwen3.5 397B leads
ResearchClawBenchSource 14.0%14.2%Qwen3.5 397B leads
JobBenchSource 8.7%—Not comparable
AA Agentic IndexSource 21.7%19.9%Kimi K2.5 leads
GDPval-AASource 25.4%23.1%Kimi K2.5 leads
GDPval-AASource 1009962Kimi K2.5 leads
VITA-BenchSource —43.7%Not comparable
CodingQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
SWE-bench VerifiedSource 76.8%76.2%Kimi K2.5 leads
SWE-bench Verified*Source 70.8%—Not comparable
LiveCodeBench v6Source 85.0%83.6%Kimi K2.5 leads
SWE-bench ProSource 50.7%50.9%Qwen3.5 397B leads
SWE MultilingualSource 73%—Not comparable
SWE-RebenchSource 58.5%—Not comparable
React Native EvalsSource 77.2%—Not comparable
SciCodeSource 48.7%—Not comparable
AA-SciCodeSource 49.0%42.0%Kimi K2.5 leads
AA Coding IndexSource 46.8%48.2%Qwen3.5 397B leads
ReasoningQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
LongBench v2Source 61%63.2%Qwen3.5 397B leads
AA-LCRSource 65.3%65.7%Qwen3.5 397B leads
CritPtSource 3.1%1.7%Kimi K2.5 leads
AI-NeedleSource —68.7%Not comparable
KnowledgeKimi K2.5 wins
BenchmarkKimi K2.5Qwen3.5 397BResult
GPQASource 87.6%88.4%Qwen3.5 397B leads
GPQA-DSource 87.6%—Not comparable
SuperGPQASource 69.2%70.4%Qwen3.5 397B leads
MMLU-ProSource 87.1%87.8%Qwen3.5 397B leads
MMLU-Pro (Arcee)Source 87.1%—Not comparable
HLESource 30.1%28.7%Kimi K2.5 leads
Artificial Analysis Intelligence IndexSource 35.4%33.7%Kimi K2.5 leads
AA-GPQA DiamondSource 87.9%89.3%Qwen3.5 397B leads
AA-HLESource 29.4%27.3%Kimi K2.5 leads
AA-Omniscience IndexSource -8.1%-29.8%Kimi K2.5 leads
AA-Omniscience AccuracySource 34.3%31.4%Kimi K2.5 leads
AA-Omniscience Hallucination RateSource 64.6%89.1%Kimi K2.5 leads
MMLU-ReduxSource —94.9%Not comparable
C-EvalSource —93%Not comparable
MathQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
AIME 2025Source 96.1%—Not comparable
AIME26Source 95.8%93.3%Kimi K2.5 leads
AIME25 (Arcee)Source 96.3%—Not comparable
HMMT Feb 2025Source 95.4%94.8%Kimi K2.5 leads
HMMT Nov 2025Source 91.1%92.7%Qwen3.5 397B leads
HMMT Feb 2026Source 87.1%87.9%Qwen3.5 397B leads
MMAnswerBenchSource 81.8%80.9%Kimi K2.5 leads
FrontierMath v2 (Tiers 1-3)Source 27.900%—Not comparable
FrontierMath v2 (Tier 4)Source 4.200%—Not comparable
MultilingualQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
MMLU-ProXSource 82.3%84.7%Qwen3.5 397B leads
NOVA-63Source 56.0%59.1%Qwen3.5 397B leads
MultimodalQwen3.5 397B wins
BenchmarkKimi K2.5Qwen3.5 397BResult
MMMU-ProSource 78.5%79%Qwen3.5 397B leads
Video-MMESource 87.4%—Not comparable
MMVUSource 80.4%—Not comparable
VideoMMMUSource 86.6%84.7%Kimi K2.5 leads
AA-MMMU-ProSource 75.4%77.3%Qwen3.5 397B leads
Design Arena WebsiteSource 1282—Not comparable
MathVisionSource —88.6%Not comparable
CharXivSource —80.8%Not comparable
ScreenSpot ProSource —65.6%Not comparable
V*Source —95.8%Not comparable
Inst. FollowingKimi K2.5 wins
BenchmarkKimi K2.5Qwen3.5 397BResult
IFEvalSource 93.9%92.6%Kimi K2.5 leads
AA-IFBenchSource 70.2%78.8%Qwen3.5 397B leads
Frequently Asked Questions (9)

Which is better, Kimi K2.5 or Qwen3.5 397B?

Kimi K2.5 is ahead on BenchLM's BenchAlign leaderboard, 59.66 to 57.01. The biggest single separator in this matchup is AIME26, where the scores are 95.8% and 93.3%.

Which is better for knowledge tasks, Kimi K2.5 or Qwen3.5 397B?

Kimi K2.5 has the edge for knowledge tasks in this comparison, averaging 56.9 versus 56.6. 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 Qwen3.5 397B?

Qwen3.5 397B has the edge for coding in this comparison, averaging 66.5 versus 59.4. Inside this category, AA-SciCode is the benchmark that creates the most daylight between them.

Which is better for math, Kimi K2.5 or Qwen3.5 397B?

Qwen3.5 397B has the edge for math in this comparison, averaging 90.6 versus 60.6. Inside this category, AIME26 is the benchmark that creates the most daylight between them.

Which is better for reasoning, Kimi K2.5 or Qwen3.5 397B?

Qwen3.5 397B has the edge for reasoning in this comparison, averaging 63.2 versus 61. Inside this category, LongBench v2 is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Kimi K2.5 or Qwen3.5 397B?

Qwen3.5 397B has the edge for agentic tasks in this comparison, averaging 56.5 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 Qwen3.5 397B?

Qwen3.5 397B has the edge for multimodal and grounded tasks in this comparison, averaging 79.6 versus 78.5. Inside this category, VideoMMMU is the benchmark that creates the most daylight between them.

Which is better for instruction following, Kimi K2.5 or Qwen3.5 397B?

Kimi K2.5 has the edge for instruction following in this comparison, averaging 93.9 versus 92.6. 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 Qwen3.5 397B?

Qwen3.5 397B has the edge for multilingual tasks in this comparison, averaging 84.7 versus 82.3. Inside this category, NOVA-63 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
Qwen3.5 397B
API / mo$3,150
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 20, 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.