VerdictPal
Join

Models / Moonshot

Kimi K2.5

Kimi · Released 2026-01-27

Moonshot's January 2026 multimodal open-weights model. 262K context, $0.60/$3 per 1M tokens. Agent Swarm up to 100 sub-agents. Predecessor to K2.6.

Open-weights Moonshot pick when K2.6 isn't available or you need the K2.5 weight checkpoint.

#31 of 51 on AA Index · current snapshot
35 AA Index · 13 benchmark rows · 4 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite53
Context262K
Input / 1M$0.60
Output / 1M$3.0
Knowledge cutoff2025-10-01
Output speed50 t/s
TTFT (default API)1.02s
Statussolid
KnowledgeReasoningCodingAgenticLong context
Knowledge
86
Reasoning
88
Coding
49
Agentic
46
Long context
65

67average across 5 tested families

Benchmark placements

Where Kimi K2.5 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
LongBench v2Long context#15/ 1948LongBench v2 leaderboard2026-04-01
LiveBenchReasoning#16/ 2161LiveBench leaderboard2026-04-15
MMLUKnowledge#18/ 2286Moonshot Kimi K2.52026-01-27
SciCodeCoding#31/ 6549Artificial Analysis2026-07-21
GPQA DiamondReasoning#32/ 6988Artificial Analysis2026-07-21
IFBenchReasoning#35/ 5770Artificial Analysis2026-07-21
Humanity's Last ExamReasoning#37/ 6529Artificial Analysis2026-07-21
AA-LCRLong context#39/ 6465Artificial Analysis2026-07-21
AA time to first tokenPerformance#43/ 69Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#44/ 6735Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#48/ 6414Artificial Analysis2026-07-21
Terminal-BenchAgentic#48/ 6946Artificial Analysis2026-07-21
AA output speedPerformance#55/ 69Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Kimi K2.5 shows up, it's highlighted.

KnowledgeKimi K2.5 · 86
  1. 1
    GPT 5.5
    91
  2. 2
    Claude Opus 4.8
    91
  3. 3
    Claude Sonnet 4.6
    91
ReasoningKimi K2.5 · 88
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingKimi K2.5 · 49
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticKimi K2.5 · 46
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextKimi K2.5 · 65
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Newest receipts

Benchmark rows added to the public ledger for Kimi K2.5 in the last 120 days. Older rows live in the full table below.

12 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context65.33%Artificial Analysis2
AA output speedPerformance50 t/sArtificial Analysis2
AA time to first tokenPerformance1.02sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning35.4%Artificial Analysis2
GPQA DiamondReasoning87.9%Artificial Analysis2
Humanity's Last ExamReasoning29.4%Artificial Analysis2
IFBenchReasoning70.2%Artificial Analysis2
SciCodeCoding49%Artificial Analysis2
τ³-Bench BankingAgentic14.23%Artificial Analysis2
Terminal-BenchAgentic45.69%Artificial Analysis2
LiveBenchReasoning61.2%LiveBench leaderboard99
LongBench v2Long context48.2%LongBench v2 leaderboard113
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Knowledge86
Reasoning88
Coding49
Agentic46
Long context65
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Kimi K2.5. Scores link to the original source; gaps mean no public row exists yet.

13 sourced rows

13 of 32 catalog benchmarks have a sourced row for Kimi K2.5.

Knowledge

Reasoning

Coding

Agentic

Long context

Performance

Sources