Skip to main content
BenchLM
Advertise

Model comparison

DeepSeek V4 Pro Base vs Kimi K2.5

Data verified

Head-to-head evidence from 3 shared benchmark results across 2 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

N/A
No comparison
Kimi K2.5
Moonshot AI
59.66/100
1 category wins1 category wins

Public leaderboard positions: DeepSeek V4 Pro Base unranked (Not scored); Kimi K2.5 #54 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. DeepSeek V4 Pro Base and Kimi K2.5 share 3 comparable benchmark results. 2 of 8 categories are comparable. 21 results are unique to DeepSeek V4 Pro Base; 60 to Kimi K2.5.

Updated July 23, 2026
Shared results
3
DeepSeek V4 Pro Base only
21
Kimi K2.5 only
60
Comparable categories
2 / 8

Treat this as a split decision. DeepSeek V4 Pro Base makes more sense if knowledge is the priority or you need the larger 1M context window; Kimi K2.5 is the better fit if reasoning is the priority.

Confidence note. This is a partial-evidence comparison with 3 shared benchmark results across 2 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

DeepSeek V4 Pro Base and Kimi K2.5 finish on the same BenchAlign overall score, so this is less about a single winner and more about where the edge shows up. The BenchAlign headline says tie; the benchmark table is where the real choice happens.

DeepSeek V4 Pro Base 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 DeepSeek V4 Pro Base and Kimi K2.5
CategoryDeepSeek V4 Pro BaseΔKimi K2.5
KnowledgeDeepSeek V4 Pro Base66.4Margin← 9.5Kimi K2.556.9
ReasoningDeepSeek V4 Pro Base51.5Margin→ 9.5Kimi K2.561.0
AgenticDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.555.0
CodingDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.559.4
MathDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.560.6
MultilingualDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.582.3
MultimodalDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.578.5
Inst. FollowingDeepSeek V4 Pro BaseNot measuredMarginNo overlapKimi K2.593.9

Decisive benchmark drivers

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

More
A · DeepSeek V4 Pro BaseB · Kimi K2.5

  1. SuperGPQA

    Knowledge
    Source ↗
    A 53.9%B 69.2%
    Winner: Kimi K2.5Δ 15.3
    SuperGPQA: DeepSeek V4 Pro Base scored 53.9%; Kimi K2.5 scored 69.2%. Kimi K2.5 wins this benchmark.

  2. MMLU-Pro

    Knowledge
    Source ↗
    A 73.5%B 87.1%
    Winner: Kimi K2.5Δ 13.6
    MMLU-Pro: DeepSeek V4 Pro Base scored 73.5%; Kimi K2.5 scored 87.1%. Kimi K2.5 wins this benchmark.

  3. LongBench v2

    Reasoning
    Source ↗
    A 51.5%B 61%
    Winner: Kimi K2.5Δ 9.5
    LongBench v2: DeepSeek V4 Pro Base scored 51.5%; Kimi K2.5 scored 61%. Kimi K2.5 wins this benchmark.

Operational comparison

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

MetricDeepSeek V4 Pro BaseKimi K2.5Comparison
Input / output priceUSD per 1M tokensDeepSeek V4 Pro BaseNot availableKimi K2.5$0.6 input / $3 outputA complete price comparison is not available.
Generation speedtokens per secondDeepSeek V4 Pro BaseNot availableKimi K2.545 tok/sA complete speed comparison is not available.
First-answer latencyseconds to first tokenDeepSeek V4 Pro BaseNot availableKimi K2.52.38 sA complete latency comparison is not available.
Context windowmaximum listed tokensDeepSeek V4 Pro Base1MKimi K2.5256KDeepSeek V4 Pro Base lists the larger context window.

Benchmark Deep Dive

Agentic
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
Terminal-Bench 2.0Source —50.8%Not comparable
BrowseCompSource —60.6%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
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%Not comparable
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
Coding
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
BigCodeBenchSource 59.2%—Not comparable
HumanEvalSource 76.8%—Not comparable
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%Not comparable
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%Not comparable
AA Coding IndexSource —46.8%Not comparable
ReasoningKimi K2.5 wins
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
BBHSource 87.5%—Not comparable
DROPSource 88.7%—Not comparable
HellaSwagSource 88.0%—Not comparable
WinoGrandeSource 81.5%—Not comparable
CLUEWSCSource 85.2%—Not comparable
LongBench v2Source 51.5%61%Kimi K2.5 leads
AA-LCRSource —65.3%Not comparable
CritPtSource —3.1%Not comparable
KnowledgeDeepSeek V4 Pro Base wins
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
AGIEvalSource 83.1%—Not comparable
MMLUSource 90.1%—Not comparable
MMLU-ReduxSource 90.8%—Not comparable
MMLU-ProSource 73.5%87.1%Kimi K2.5 leads
MMMLUSource 90.3%—Not comparable
C-EvalSource 93.1%—Not comparable
CMMLUSource 90.8%—Not comparable
MultiLoKoSource 51.1%—Not comparable
SimpleQASource 55.2%—Not comparable
SuperGPQASource 53.9%69.2%Kimi K2.5 leads
FACTS ParametricSource 62.6%—Not comparable
TriviaQASource 85.6%—Not comparable
GPQASource —87.6%Not comparable
GPQA-DSource —87.6%Not comparable
MMLU-Pro (Arcee)Source —87.1%Not comparable
HLESource —30.1%Not comparable
Artificial Analysis Intelligence IndexSource —35.4%Not comparable
AA-GPQA DiamondSource —87.9%Not comparable
AA-HLESource —29.4%Not comparable
AA-Omniscience IndexSource —-8.1%Not comparable
AA-Omniscience AccuracySource —34.3%Not comparable
AA-Omniscience Hallucination RateSource —64.6%Not comparable
Math
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
GSM8KSource 92.6%—Not comparable
MATHSource 64.5%—Not comparable
CMathSource 90.9%—Not comparable
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%Not comparable
FrontierMath v2 (Tier 4)Source —4.200%Not comparable
Multilingual
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
MGSMSource 84.4%—Not comparable
MMLU-ProXSource —82.3%Not comparable
NOVA-63Source —56.0%Not comparable
Multimodal
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
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
BenchmarkDeepSeek V4 Pro BaseKimi K2.5Result
IFEvalSource —93.9%Not comparable
AA-IFBenchSource —70.2%Not comparable
Frequently Asked Questions (3)

Which is better, DeepSeek V4 Pro Base or Kimi K2.5?

DeepSeek V4 Pro Base and Kimi K2.5 are tied on the BenchAlign overall score, so the right pick depends on which category matters most for your use case.

Which is better for knowledge tasks, DeepSeek V4 Pro Base or Kimi K2.5?

DeepSeek V4 Pro Base has the edge for knowledge tasks in this comparison, averaging 66.4 versus 56.9. Inside this category, SuperGPQA is the benchmark that creates the most daylight between them.

Which is better for reasoning, DeepSeek V4 Pro Base or Kimi K2.5?

Kimi K2.5 has the edge for reasoning in this comparison, averaging 61 versus 51.5. Inside this category, LongBench v2 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.

DeepSeek V4 Pro Base
API / mo$0
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

Explore More

DeepSeek 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.