Skip to main content

Model comparison

GPT-5.5 vs Kimi K2.5

Data verified

Head-to-head evidence from 33 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

OpenAI
73.51/100
Margin
13.9pts
← winning
Moonshot AI
59.66/100
3 category wins3 category wins

Public leaderboard positions: GPT-5.5 #9 (Estimated); Kimi K2.5 #54 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. GPT-5.5 and Kimi K2.5 share 33 comparable benchmark results. 6 of 8 categories are comparable. 24 results are unique to GPT-5.5; 30 to Kimi K2.5.

Updated July 22, 2026
Shared results
33
GPT-5.5 only
24
Kimi K2.5 only
30
Comparable categories
6 / 8

Pick GPT-5.5 if you want the stronger benchmark profile. Kimi K2.5 only becomes the better choice if mathematics is the priority or you want the cheaper token bill.

Confidence note. This is a partial-evidence comparison with 33 shared benchmark results across 7 evidence categories; 6 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

GPT-5.5 is clearly ahead on the BenchAlign aggregate, 73.51 to 59.66. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

GPT-5.5's sharpest advantage is in agentic, where it averages 81.6 against 55. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 82% to 50.8%. Kimi K2.5 does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.

GPT-5.5 is also the more expensive model on tokens at $5.00 input / $30.00 output per 1M tokens, versus $0.60 input / $3.00 output per 1M tokens for Kimi K2.5. That is roughly 10.0x on output cost alone. GPT-5.5 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. GPT-5.5 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 GPT-5.5 and Kimi K2.5
CategoryGPT-5.5ΔKimi K2.5
AgenticGPT-5.581.6Margin 26.6Kimi K2.555.0
ReasoningGPT-5.585.0Margin 24.0Kimi K2.561.0
MathGPT-5.547.6Margin 13.0Kimi K2.560.6
MultimodalGPT-5.570.4Margin 8.1Kimi K2.578.5
KnowledgeGPT-5.557.8Margin 0.9Kimi K2.556.9
CodingGPT-5.558.6Margin 0.8Kimi K2.559.4
MultilingualGPT-5.5Not measuredMarginNo overlapKimi K2.582.3
Inst. FollowingGPT-5.5Not measuredMarginNo overlapKimi K2.593.9

Decisive benchmark drivers

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

More
A · GPT-5.5B · Kimi K2.5
  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 82%B 50.8%
    Winner: GPT-5.5Δ 31.2
    Terminal-Bench 2.0: GPT-5.5 scored 82%; Kimi K2.5 scored 50.8%. GPT-5.5 wins this benchmark.
  2. FrontierMath v2 (Tier 4)

    Math
    Source ↗
    A 35.400%B 4.200%
    Winner: GPT-5.5Δ 31.2
    FrontierMath v2 (Tier 4): GPT-5.5 scored 35.400%; Kimi K2.5 scored 4.200%. GPT-5.5 wins this benchmark.
  3. BrowseComp

    Agentic
    Source ↗
    A 84.4%B 60.6%
    Winner: GPT-5.5Δ 23.8
    BrowseComp: GPT-5.5 scored 84.4%; Kimi K2.5 scored 60.6%. GPT-5.5 wins this benchmark.
  4. FrontierMath v2 (Tiers 1-3)

    Math
    Source ↗
    A 51.700%B 27.900%
    Winner: GPT-5.5Δ 23.8
    FrontierMath v2 (Tiers 1-3): GPT-5.5 scored 51.700%; Kimi K2.5 scored 27.900%. GPT-5.5 wins this benchmark.
  5. HLE

    Knowledge
    Source ↗
    A 52.2%B 30.1%
    Winner: GPT-5.5Δ 22.1
    HLE: GPT-5.5 scored 52.2%; Kimi K2.5 scored 30.1%. GPT-5.5 wins this benchmark.

Operational comparison

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

MetricGPT-5.5Kimi K2.5Comparison
Input / output priceUSD per 1M tokensGPT-5.5$5 input / $30 outputKimi K2.5$0.6 input / $3 outputKimi K2.5 has the lower combined listed price.
Generation speedtokens per secondGPT-5.5Not availableKimi K2.545 tok/sA complete speed comparison is not available.
First-answer latencyseconds to first tokenGPT-5.5Not availableKimi K2.52.38 sA complete latency comparison is not available.
Context windowmaximum listed tokensGPT-5.51MKimi K2.5256KGPT-5.5 lists the larger context window.

Benchmark Deep Dive

AgenticGPT-5.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
Terminal-Bench 2.0Source 82%50.8%GPT-5.5 leads
CyberGymSource 81.8%Not comparable
BrowseCompSource 84.4%60.6%GPT-5.5 leads
OSWorld-VerifiedSource 78.7%Not comparable
MCP AtlasSource 75.3%29.5%GPT-5.5 leads
ToolathlonSource 55.6%27.8%GPT-5.5 leads
τ²-bench resultsSource 93.9%95.9%Kimi K2.5 leads
AA Agentic IndexSource 44.9%21.7%GPT-5.5 leads
APEX-Agents-AASource 37.7%11.5%GPT-5.5 leads
GDPval-AASource 49.5%25.4%GPT-5.5 leads
GDPval-AASource 14901009GPT-5.5 leads
Gert LabsSource 72.93%45.88%GPT-5.5 leads
ResearchClawBenchSource 17.0%14.0%GPT-5.5 leads
OSWorld 2.0Source 13.0%Not comparable
JobBenchSource 42.7%8.7%GPT-5.5 leads
ExploitGymSource 13.4%Not comparable
AA BriefcaseSource 1154Not comparable
AA AutomationBenchSource 42.1%Not comparable
AA EnterpriseOps-GymSource 46.6%Not comparable
AA Harvey LABSource 86.3%Not comparable
AA ITBenchSource 45.8%Not comparable
AA Tau3 BankingSource 31.3%Not comparable
terminalBenchHardSource 60.6%Not comparable
aaTerminalBench21Source 84.3%Not comparable
Claw-EvalSource 52.3%Not comparable
QwenClawBenchSource 54.3%Not comparable
τ³-bench resultsSource 65.7%Not comparable
DeepSearchQASource 77.1%Not comparable
DeepPlanningSource 14.4%Not comparable
MCP-TasksSource 59.1%Not comparable
WideResearchSource 72.7%Not comparable
CodingKimi K2.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
SWE-bench ProSource 58.6%50.7%GPT-5.5 leads
Terminal-Bench 2.0Source 82.0%Not comparable
Vibe Code BenchSource 69.85%Not comparable
React Native EvalsSource 84.7%77.2%GPT-5.5 leads
cursorBench31Source 59.2%Not comparable
cursorBench32Source 58.4%Not comparable
AA Coding IndexSource 74.9%46.8%GPT-5.5 leads
AA-SciCodeSource 56.1%49.0%GPT-5.5 leads
FrontierCode 1.1 MainSource 43.0%Not comparable
SWE-bench VerifiedSource 76.8%Not comparable
SWE-bench Verified*Source 70.8%Not comparable
LiveCodeBench v6Source 85.0%Not comparable
SWE MultilingualSource 73%Not comparable
SWE-RebenchSource 58.5%Not comparable
SciCodeSource 48.7%Not comparable
ReasoningGPT-5.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
MRCR v2 64K-128KSource 83.1%Not comparable
MRCR v2 128K-256KSource 87.5%Not comparable
ARC-AGI-2Source 85%Not comparable
AA-LCRSource 74.3%65.3%GPT-5.5 leads
CritPtSource 27.1%3.1%GPT-5.5 leads
LongBench v2Source 61%Not comparable
KnowledgeGPT-5.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
GPQASource 93.6%87.6%GPT-5.5 leads
GPQA-DSource 93.6%87.6%GPT-5.5 leads
HLESource 52.2%30.1%GPT-5.5 leads
HLE w/o toolsSource 41.4%Not comparable
Artificial Analysis Intelligence IndexSource 54.8%35.4%GPT-5.5 leads
AA-GPQA DiamondSource 93.5%87.9%GPT-5.5 leads
AA-HLESource 44.3%29.4%GPT-5.5 leads
AA-Omniscience IndexSource 20.1%-8.1%GPT-5.5 leads
AA-Omniscience AccuracySource 56.9%34.3%GPT-5.5 leads
AA-Omniscience Hallucination RateSource 85.5%64.6%Kimi K2.5 leads
SuperGPQASource 69.2%Not comparable
MMLU-ProSource 87.1%Not comparable
MMLU-Pro (Arcee)Source 87.1%Not comparable
MathKimi K2.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
FrontierMath (legacy)Source 51.7%Not comparable
FrontierMath v2 (Tiers 1-3)Source 51.700%27.900%GPT-5.5 leads
FrontierMath v2 (Tier 4)Source 35.400%4.200%GPT-5.5 leads
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
Multilingual
BenchmarkGPT-5.5Kimi K2.5Result
MMLU-ProXSource 82.3%Not comparable
NOVA-63Source 56.0%Not comparable
MultimodalKimi K2.5 wins
BenchmarkGPT-5.5Kimi K2.5Result
MMMU-ProSource 81.2%78.5%GPT-5.5 leads
MMMU-Pro w/ PythonSource 83.2%Not comparable
OfficeQA ProSource 54.1%Not comparable
AA-MMMU-ProSource 79.9%75.4%GPT-5.5 leads
Design Arena WebsiteSource 12821279GPT-5.5 leads
Video-MMESource 87.4%Not comparable
MMVUSource 80.4%Not comparable
VideoMMMUSource 86.6%Not comparable
Inst. Following
BenchmarkGPT-5.5Kimi K2.5Result
AA-IFBenchSource 75.9%70.2%GPT-5.5 leads
IFEvalSource 93.9%Not comparable
Frequently Asked Questions (7)

Which is better, GPT-5.5 or Kimi K2.5?

GPT-5.5 is ahead on BenchLM's BenchAlign leaderboard, 73.51 to 59.66. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 82% and 50.8%.

Which is better for knowledge tasks, GPT-5.5 or Kimi K2.5?

GPT-5.5 has the edge for knowledge tasks in this comparison, averaging 57.8 versus 56.9. Inside this category, AA-Omniscience Index is the benchmark that creates the most daylight between them.

Which is better for coding, GPT-5.5 or Kimi K2.5?

Kimi K2.5 has the edge for coding in this comparison, averaging 59.4 versus 58.6. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.

Which is better for math, GPT-5.5 or Kimi K2.5?

Kimi K2.5 has the edge for math in this comparison, averaging 60.6 versus 47.6. Inside this category, FrontierMath v2 (Tier 4) is the benchmark that creates the most daylight between them.

Which is better for reasoning, GPT-5.5 or Kimi K2.5?

GPT-5.5 has the edge for reasoning in this comparison, averaging 85 versus 61. Inside this category, CritPt is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, GPT-5.5 or Kimi K2.5?

GPT-5.5 has the edge for agentic tasks in this comparison, averaging 81.6 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, GPT-5.5 or Kimi K2.5?

Kimi K2.5 has the edge for multimodal and grounded tasks in this comparison, averaging 78.5 versus 70.4. Inside this category, AA-MMMU-Pro 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.

GPT-5.5
API / mo$26,250
Self-host / moNot listed
Break-even
Proprietary model — self-hosting not applicable.
Kimi K2.5
API / mo$2,700
Self-host / mo$5,221
Break-even132M/day
Model the full break-even

Related Comparisons

Last updated: July 22, 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.