Skip to main content

Model comparison

Kimi K2.5 vs Qwen 3.6 Max (preview)

Data verified

Head-to-head evidence from 18 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

Moonshot AI
59.66/100
Margin
0.1pts
winning →
59.72/100
2 category wins2 category wins

Public leaderboard positions: Kimi K2.5 #54 (Supported); Qwen 3.6 Max (preview) #53 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Kimi K2.5 and Qwen 3.6 Max (preview) share 18 comparable benchmark results. 4 of 8 categories are comparable. 45 results are unique to Kimi K2.5; 3 to Qwen 3.6 Max (preview).

Updated July 23, 2026
Shared results
18
Kimi K2.5 only
45
Qwen 3.6 Max (preview) only
3
Comparable categories
4 / 8

Pick Qwen 3.6 Max (preview) if you want the stronger benchmark profile. Kimi K2.5 only becomes the better choice if mathematics 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 18 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

Qwen 3.6 Max (preview) has the cleaner BenchAlign overall profile here, landing at 59.72 versus 59.66. It is a real lead, but still close enough that category-level strengths matter more than the headline number.

Qwen 3.6 Max (preview)'s sharpest advantage is in knowledge, where it averages 73.9 against 56.9. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 50.8% to 65.4%. Kimi K2.5 does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.

Qwen 3.6 Max (preview) 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.

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 Qwen 3.6 Max (preview)
CategoryKimi K2.5ΔQwen 3.6 Max (preview)
MathKimi K2.560.6Margin 42.2Qwen 3.6 Max (preview)18.4
KnowledgeKimi K2.556.9Margin 17.0Qwen 3.6 Max (preview)73.9
AgenticKimi K2.555.0Margin 10.4Qwen 3.6 Max (preview)65.4
CodingKimi K2.559.4Margin 8.4Qwen 3.6 Max (preview)51.0
ReasoningKimi K2.561.0MarginNo overlapQwen 3.6 Max (preview)Not measured
MultilingualKimi K2.582.3MarginNo overlapQwen 3.6 Max (preview)Not measured
MultimodalKimi K2.578.5MarginNo overlapQwen 3.6 Max (preview)Not measured
Inst. FollowingKimi K2.593.9MarginNo overlapQwen 3.6 Max (preview)Not measured

Decisive benchmark drivers

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

More
A · Kimi K2.5B · Qwen 3.6 Max (preview)
  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 50.8%B 65.4%
    Winner: Qwen 3.6 Max (preview)Δ 14.6
    Terminal-Bench 2.0: Kimi K2.5 scored 50.8%; Qwen 3.6 Max (preview) scored 65.4%. Qwen 3.6 Max (preview) wins this benchmark.
  2. SWE-bench Pro

    Coding
    Source ↗
    A 50.7%B 57.3%
    Winner: Qwen 3.6 Max (preview)Δ 6.6
    SWE-bench Pro: Kimi K2.5 scored 50.7%; Qwen 3.6 Max (preview) scored 57.3%. Qwen 3.6 Max (preview) wins this benchmark.
  3. FrontierMath v2 (Tiers 1-3)

    Math
    Source ↗
    A 27.900%B 23.103%
    Winner: Kimi K2.5Δ 4.8
    FrontierMath v2 (Tiers 1-3): Kimi K2.5 scored 27.900%; Qwen 3.6 Max (preview) scored 23.103%. Kimi K2.5 wins this benchmark.
  4. SuperGPQA

    Knowledge
    Source ↗
    A 69.2%B 73.9%
    Winner: Qwen 3.6 Max (preview)Δ 4.7
    SuperGPQA: Kimi K2.5 scored 69.2%; Qwen 3.6 Max (preview) scored 73.9%. Qwen 3.6 Max (preview) wins this benchmark.
  5. SciCode

    Coding
    Source ↗
    A 48.7%B 47%
    Winner: Kimi K2.5Δ 1.7
    SciCode: Kimi K2.5 scored 48.7%; Qwen 3.6 Max (preview) scored 47%. Kimi K2.5 wins this benchmark.

Operational comparison

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

MetricKimi K2.5Qwen 3.6 Max (preview)Comparison
Input / output priceUSD per 1M tokensKimi K2.5$0.6 input / $3 outputQwen 3.6 Max (preview)Not availableA complete price comparison is not available.
Generation speedtokens per secondKimi K2.545 tok/sQwen 3.6 Max (preview)Not availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenKimi K2.52.38 sQwen 3.6 Max (preview)Not availableA complete latency comparison is not available.
Context windowmaximum listed tokensKimi K2.5256KQwen 3.6 Max (preview)256KListed context windows are equal.

Benchmark Deep Dive

AgenticQwen 3.6 Max (preview) wins
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
Terminal-Bench 2.0Source 50.8%65.4%Qwen 3.6 Max (preview) leads
BrowseCompSource 60.6%Not comparable
Claw-EvalSource 52.3%Not comparable
QwenClawBenchSource 54.3%59.0%Qwen 3.6 Max (preview) 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%Not comparable
MCP-TasksSource 59.1%Not comparable
WideResearchSource 72.7%Not comparable
τ²-bench resultsSource 95.9%95.9%Tie
APEX-Agents-AASource 11.5%Not comparable
Gert LabsSource 45.88%Not comparable
ResearchClawBenchSource 14.0%Not comparable
JobBenchSource 8.7%Not comparable
AA Agentic IndexSource 21.7%Not comparable
GDPval-AASource 25.4%Not comparable
GDPval-AASource 1009Not comparable
QwenWebBenchSource 1532Not comparable
CodingKimi K2.5 wins
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
SWE-bench VerifiedSource 76.8%Not comparable
SWE-bench Verified*Source 70.8%Not comparable
LiveCodeBench v6Source 85.0%Not comparable
SWE-bench ProSource 50.7%57.3%Qwen 3.6 Max (preview) leads
SWE MultilingualSource 73%Not comparable
SWE-RebenchSource 58.5%Not comparable
React Native EvalsSource 77.2%Not comparable
SciCodeSource 48.7%47%Kimi K2.5 leads
AA-SciCodeSource 49.0%46.9%Kimi K2.5 leads
AA Coding IndexSource 46.8%Not comparable
NL2RepoSource 42.9%Not comparable
Terminal-Bench 2.0Source 65.4%Not comparable
Reasoning
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
LongBench v2Source 61%Not comparable
AA-LCRSource 65.3%69.7%Qwen 3.6 Max (preview) leads
CritPtSource 3.1%3.7%Qwen 3.6 Max (preview) leads
KnowledgeQwen 3.6 Max (preview) wins
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
GPQASource 87.6%Not comparable
GPQA-DSource 87.6%Not comparable
SuperGPQASource 69.2%73.9%Qwen 3.6 Max (preview) leads
MMLU-ProSource 87.1%Not comparable
MMLU-Pro (Arcee)Source 87.1%Not comparable
HLESource 30.1%Not comparable
Artificial Analysis Intelligence IndexSource 35.4%40.0%Qwen 3.6 Max (preview) leads
AA-GPQA DiamondSource 87.9%88.8%Qwen 3.6 Max (preview) leads
AA-HLESource 29.4%28.9%Kimi K2.5 leads
AA-Omniscience IndexSource -8.1%10.2%Qwen 3.6 Max (preview) leads
AA-Omniscience AccuracySource 34.3%37.7%Qwen 3.6 Max (preview) leads
AA-Omniscience Hallucination RateSource 64.6%44.2%Qwen 3.6 Max (preview) leads
MathKimi K2.5 wins
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
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%Not comparable
MMAnswerBenchSource 81.8%Not comparable
FrontierMath v2 (Tiers 1-3)Source 27.900%23.103%Kimi K2.5 leads
FrontierMath v2 (Tier 4)Source 4.200%4.167%Kimi K2.5 leads
Multilingual
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
MMLU-ProXSource 82.3%Not comparable
NOVA-63Source 56.0%Not comparable
Multimodal
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
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 1279Not comparable
Inst. Following
BenchmarkKimi K2.5Qwen 3.6 Max (preview)Result
IFEvalSource 93.9%Not comparable
AA-IFBenchSource 70.2%76.6%Qwen 3.6 Max (preview) leads
Frequently Asked Questions (5)

Which is better, Kimi K2.5 or Qwen 3.6 Max (preview)?

Qwen 3.6 Max (preview) is ahead on BenchLM's BenchAlign leaderboard, 59.72 to 59.66. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 50.8% and 65.4%.

Which is better for knowledge tasks, Kimi K2.5 or Qwen 3.6 Max (preview)?

Qwen 3.6 Max (preview) has the edge for knowledge tasks in this comparison, averaging 73.9 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 Qwen 3.6 Max (preview)?

Kimi K2.5 has the edge for coding in this comparison, averaging 59.4 versus 51. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.

Which is better for math, Kimi K2.5 or Qwen 3.6 Max (preview)?

Kimi K2.5 has the edge for math in this comparison, averaging 60.6 versus 18.4. Inside this category, FrontierMath v2 (Tiers 1-3) is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Kimi K2.5 or Qwen 3.6 Max (preview)?

Qwen 3.6 Max (preview) has the edge for agentic tasks in this comparison, averaging 65.4 versus 55. Inside this category, Terminal-Bench 2.0 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
Qwen 3.6 Max (preview)
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.