Skip to main content

Model comparison

Kimi K2.5 (Reasoning) vs Qwen3.5 397B

Data verified

Head-to-head evidence from 24 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

59.35/100
Margin
2.3pts
← winning
57.01/100
2 category wins2 category wins

Public leaderboard positions: Kimi K2.5 (Reasoning) #57 (Estimated); Qwen3.5 397B #71 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Kimi K2.5 (Reasoning) and Qwen3.5 397B share 24 comparable benchmark results. 4 of 8 categories are comparable. 3 results are unique to Kimi K2.5 (Reasoning); 31 to Qwen3.5 397B.

Updated July 20, 2026
Shared results
24
Kimi K2.5 (Reasoning) only
3
Qwen3.5 397B only
31
Comparable categories
4 / 8

Pick Kimi K2.5 (Reasoning) if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if agentic 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 24 shared benchmark results across 6 evidence categories; 4 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Kimi K2.5 (Reasoning) has the cleaner BenchAlign overall profile here, landing at 59.35 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 (Reasoning)'s sharpest advantage is in knowledge, where it averages 87.2 against 56.6. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 50.8% to 52.5%. Qwen3.5 397B does hit back in agentic, 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 (Reasoning). Kimi K2.5 (Reasoning) is the reasoning model in the pair, while Qwen3.5 397B 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.

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 (Reasoning) and Qwen3.5 397B
CategoryKimi K2.5 (Reasoning)ΔQwen3.5 397B
KnowledgeKimi K2.5 (Reasoning)87.2Margin 30.6Qwen3.5 397B56.6
CodingKimi K2.5 (Reasoning)76.8Margin 10.3Qwen3.5 397B66.5
AgenticKimi K2.5 (Reasoning)55.0Margin 1.5Qwen3.5 397B56.5
MultimodalKimi K2.5 (Reasoning)78.5Margin 1.1Qwen3.5 397B79.6
ReasoningKimi K2.5 (Reasoning)Not measuredMarginNo overlapQwen3.5 397B63.2
MathKimi K2.5 (Reasoning)Not measuredMarginNo overlapQwen3.5 397B90.6
MultilingualKimi K2.5 (Reasoning)Not measuredMarginNo overlapQwen3.5 397B84.7
Inst. FollowingKimi K2.5 (Reasoning)Not measuredMarginNo overlapQwen3.5 397B92.6

Decisive benchmark drivers

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

More
A · Kimi K2.5 (Reasoning)B · Qwen3.5 397B
  1. 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 (Reasoning) scored 50.8%; Qwen3.5 397B scored 52.5%. Qwen3.5 397B wins this benchmark.
  2. BrowseComp

    Agentic
    Source ↗
    A 60.6%B 62%
    Winner: Qwen3.5 397BΔ 1.4
    BrowseComp: Kimi K2.5 (Reasoning) scored 60.6%; Qwen3.5 397B scored 62%. Qwen3.5 397B wins this benchmark.
  3. GPQA

    Knowledge
    Source ↗
    A 87.6%B 88.4%
    Winner: Qwen3.5 397BΔ 0.8
    GPQA: Kimi K2.5 (Reasoning) scored 87.6%; Qwen3.5 397B scored 88.4%. Qwen3.5 397B wins this benchmark.
  4. MMLU-Pro

    Knowledge
    Source ↗
    A 87.1%B 87.8%
    Winner: Qwen3.5 397BΔ 0.7
    MMLU-Pro: Kimi K2.5 (Reasoning) scored 87.1%; Qwen3.5 397B scored 87.8%. Qwen3.5 397B wins this benchmark.
  5. SWE-bench Verified

    Coding
    Source ↗
    A 76.8%B 76.2%
    Winner: Kimi K2.5 (Reasoning)Δ 0.6
    SWE-bench Verified: Kimi K2.5 (Reasoning) scored 76.8%; Qwen3.5 397B scored 76.2%. Kimi K2.5 (Reasoning) wins this benchmark.

Operational comparison

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

MetricKimi K2.5 (Reasoning)Qwen3.5 397BComparison
Input / output priceUSD per 1M tokensKimi K2.5 (Reasoning)$0.6 input / $3 outputQwen3.5 397B$0.6 input / $3.6 outputKimi K2.5 (Reasoning) has the lower combined listed price.
Generation speedtokens per secondKimi K2.5 (Reasoning)Not availableQwen3.5 397B96 tok/sA complete speed comparison is not available.
First-answer latencyseconds to first tokenKimi K2.5 (Reasoning)Not availableQwen3.5 397B2.44 sA complete latency comparison is not available.
Context windowmaximum listed tokensKimi K2.5 (Reasoning)128KQwen3.5 397B128KListed context windows are equal.

Benchmark Deep Dive

AgenticQwen3.5 397B wins
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
Terminal-Bench 2.0Source 50.8%52.5%Qwen3.5 397B leads
BrowseCompSource 60.6%62%Qwen3.5 397B leads
APEX-Agents-AASource 11.5%15.3%Qwen3.5 397B leads
τ²-bench resultsSource 95.9%95.6%Kimi K2.5 (Reasoning) leads
Gert LabsSource 32.58%46.76%Qwen3.5 397B leads
AA Agentic IndexSource 21.7%19.9%Kimi K2.5 (Reasoning) leads
GDPval-AASource 25.4%23.1%Kimi K2.5 (Reasoning) leads
GDPval-AASource 1009962Kimi K2.5 (Reasoning) leads
Claw-EvalSource 56.8%Not comparable
QwenClawBenchSource 51.8%Not comparable
τ³-bench resultsSource 68.4%Not comparable
VITA-BenchSource 43.7%Not comparable
DeepPlanningSource 37.6%Not comparable
ToolathlonSource 36.3%Not comparable
MCP AtlasSource 46.1%Not comparable
MCP-TasksSource 74.2%Not comparable
WideResearchSource 74.0%Not comparable
ResearchClawBenchSource 14.2%Not comparable
CodingKimi K2.5 (Reasoning) wins
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
SWE-bench VerifiedSource 76.8%76.2%Kimi K2.5 (Reasoning) leads
Vibe Code BenchSource 17.54%Not comparable
AA-SciCodeSource 49.0%42.0%Kimi K2.5 (Reasoning) leads
AA Coding IndexSource 46.8%48.2%Qwen3.5 397B leads
LiveCodeBench v6Source 83.6%Not comparable
SWE-bench ProSource 50.9%Not comparable
Reasoning
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
AA-LCRSource 65.3%65.7%Qwen3.5 397B leads
CritPtSource 3.1%1.7%Kimi K2.5 (Reasoning) leads
LongBench v2Source 63.2%Not comparable
AI-NeedleSource 68.7%Not comparable
KnowledgeKimi K2.5 (Reasoning) wins
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
GPQASource 87.6%88.4%Qwen3.5 397B leads
MMLU-ProSource 87.1%87.8%Qwen3.5 397B leads
Artificial Analysis Intelligence IndexSource 35.4%33.7%Kimi K2.5 (Reasoning) leads
AA-GPQA DiamondSource 87.9%89.3%Qwen3.5 397B leads
AA-HLESource 29.4%27.3%Kimi K2.5 (Reasoning) leads
AA-Omniscience IndexSource -8.1%-29.8%Kimi K2.5 (Reasoning) leads
AA-Omniscience AccuracySource 34.3%31.4%Kimi K2.5 (Reasoning) leads
AA-Omniscience Hallucination RateSource 64.6%89.1%Kimi K2.5 (Reasoning) leads
SuperGPQASource 70.4%Not comparable
MMLU-ReduxSource 94.9%Not comparable
C-EvalSource 93%Not comparable
HLESource 28.7%Not comparable
Math
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
AIME 2025Source 96.1%Not comparable
AIME26Source 93.3%Not comparable
HMMT Feb 2025Source 94.8%Not comparable
HMMT Nov 2025Source 92.7%Not comparable
HMMT Feb 2026Source 87.9%Not comparable
MMAnswerBenchSource 80.9%Not comparable
Multilingual
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
MMLU-ProXSource 84.7%Not comparable
NOVA-63Source 59.1%Not comparable
MultimodalQwen3.5 397B wins
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
MMMU-ProSource 78.5%79%Qwen3.5 397B leads
AA-MMMU-ProSource 75.4%77.3%Qwen3.5 397B leads
Design Arena WebsiteSource 1282Not comparable
MathVisionSource 88.6%Not comparable
CharXivSource 80.8%Not comparable
VideoMMMUSource 84.7%Not comparable
ScreenSpot ProSource 65.6%Not comparable
V*Source 95.8%Not comparable
Inst. Following
BenchmarkKimi K2.5 (Reasoning)Qwen3.5 397BResult
AA-IFBenchSource 70.2%78.8%Qwen3.5 397B leads
IFEvalSource 92.6%Not comparable
Frequently Asked Questions (5)

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

Kimi K2.5 (Reasoning) is ahead on BenchLM's BenchAlign leaderboard, 59.35 to 57.01. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 50.8% and 52.5%.

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

Kimi K2.5 (Reasoning) has the edge for knowledge tasks in this comparison, averaging 87.2 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 (Reasoning) or Qwen3.5 397B?

Kimi K2.5 (Reasoning) has the edge for coding in this comparison, averaging 76.8 versus 66.5. Inside this category, AA-SciCode is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Kimi K2.5 (Reasoning) 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 (Reasoning) 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, AA-MMMU-Pro is the benchmark that creates the most daylight between them.

Related Comparisons

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.