Stable LatentMoE (2.8T total, 896 routed experts with 16 active per token; KDA/MLA hybrid attention at 3
reasoning
1 ratio)
AA Intelligence Index
reasoning
43.6
AA Coding Index
coding
76.2
AnalystAgent
strategy
38.8%
Apex Agents
agentic
41.3%
AutomationBench
agentic
58.3%
AA Briefcase
strategy
1496.6 Elo
Critpt
reasoning
23.4%
EnterpriseOps Gym
agentic
45.3%
Gdp Pdf
strategy
22.0%
AA GDPval 2.0
strategy
1583.5 Elo
GPQA Diamond
reasoning
93.5%
Harvey LAB-AA
legal
94.6%
Humanity's Last Exam (HLE)
reasoning
46.9%
Itbench
reasoning
47.7%
Lcr (1.1)
reasoning
88.7%
Mlcr
reasoning
38.3%
Mmmu Pro
reasoning
80.5%
Omniscience
science
19.7
SciCode
science
59.5%
Tau3 Banking
finance
46.0%
Terminal-Bench 2.1
coding
85.0%
Terminal-Bench 4.0
coding
12.6%
Openrouter GPQA Diamond
reasoning
91.5%
Openrouter Tau2 Bench Airline
finance
70.9%
Vals EMB 2.0
finance
66.4%
Vals Code Migration
coding
16.0%
Apprenticebench CUA
agentic
18.0%
Apprenticebench API
reasoning
25.0%
DeepSWE / SWE-bench Pro
coding
68.5%
Frontiermath Tiers 1 3 (v2)
math
72.2%
FrontierMath Tier 4
math
39.0%
SimpleQA
voice
50.6%
Lisanbench (0.2.0)
reasoning
10520.7
Chess Puzzles
reasoning
39.0%
Mystery Game Puzzles
reasoning
26.0%
Epoch GPQA Diamond
reasoning
93.1%
Frontierswe (2)
coding
25.9%
Toolathlon Verified
agentic
76.5%
Mcp Atlas
reasoning
82.3%
Context Arena MRCR V2 (8-needle)
agentic
73.3%
Mls Bench Lite (30-tasks)
reasoning
48.3
Surge Chartography (100-tasks)
reasoning
26.6%
Surge Gdp Pdf (100-tasks)
strategy
19.0%
Mcpmark (verified)
reasoning
96.1%
AA Agentic Index
agentic
50.6
TAU-Bench Banking
agentic
46.0%
Briefcase
strategy
1496.6 Elo
Gdpval (2)
strategy
1583.5 Elo
TAU-3 Banking
agentic
46.0%
TAU-2 Airline
agentic
70.9%
DeepSWE v1.1
coding
68.5%
Design Arena Frontend
coding
1284 Elo
Design Arena Fullstack
coding
1313 Elo
Briefcase v1.1
strategy
1496.6 Elo
GDPval-AA v2.1
strategy
1583.5 Elo
AA-LCR v1.1
reasoning
88.7%
AA-Omniscience Index
reasoning
19.7
SimpleQA Verified
voice
50.6%
Simple Bench
reasoning
60.7%
Arc Agi (1)
reasoning
94.5%
Arc Agi (2)
reasoning
60.4%
Furniture Assembly
finance
34.2%
Scale SWE Bench Pro V2 Full
coding
97.7%
Scale SWE Bench Pro V2 Hard
coding
88.2
Lmarena Webdev
agentic
1659.9 Elo
Lmarena Agent
agentic
4.5%
Vals Biomysterybench
healthcare
71.5%
Vals IOI
reasoning
48.9%
Terminal-Bench Science
coding
2.9%
Vals Time Horizon Index Ksp
reasoning
10.5%
Vals Vibe Code Bench 1 100
coding
18.2%
Vals Tax Agent Bench
agentic
68.7%
Architectural Profile & Capabilities
GDPval-AA v2: 1668 Elo (#3; Fable 5: 1760, Sol: 1748, Opus 4.8: 1600), AA-Briefcase: 1548 Elo (#2; Fable 5: 1583, Sol: 1495), BrowseComp: 91.2% (#1; Sol: 90.4), Terminal-Bench 2.1: 88.3% (#2; Sol: 88.8), DeepSWE: 67.5% (#3; Sol: 73.0, Fable 5: 70.0), FrontierSWE: 81.2% (#2; Fable 5: 86.6), SWE Marathon: 42.0% (#1; Opus 4.8: 40.0), Program Bench: 77.8% (#1; Sol: 77.6), Automation Bench: 30.8% (#1; Sol: 29.7), SpreadsheetBench 2: 34.8% (#1; Fable 5: 34.7), JobBench: 52.9% (#2; Fable 5: 57.4), CharXiv (RQ) w/ tool: 91.3%, Zerobench w/ tool: 41.0%, Kimi Code Bench 2.0 (internal): 72.9%. *Moonshot internal:* Online Exp: 75.5, DECK-Bench: 73.5, Finance-Bench: 62.6 (all #1 vs Opus 4.8 & GPT-5.5). *Still pending:* SWE-bench Verified, MMLU/MMLU-Pro, MATH-500 | Moonshot AI's most capable model (released Jul 16, 2026). 2.8T-parameter multimodal reasoning model — the largest in the Kimi family. 1M-token context window (4× the K2.x family's 256K) with 75% KV cache reduction and up to 6× decoding throughput at 1M context via Kimi Delta Attention (KDA). Agent Swarm scales to 300 sub-agents / 4,000 actions. Strong across coding (Terminal-Bench 88.3% #2, FrontierSWE 81.2% #2, SWE Marathon 42.0% #1), agentic automation (Automation Bench 30.8% #1, SpreadsheetBench 2 34.8% #1), long-context retrieval (BrowseComp 91.2% #1), financial/legal document analysis, patent indexing, and professional slide generation ("OK Computer"). **Best for:** long-horizon agentic coding (SWE Marathon, multi-file refactors), enterprise automation, 1M-context research/synthesis, BrowseComp-class retrieval, financial analysis, and agent-swarm coordination — comprehensive codebase reviews, multi-document financial audits, patent/legal indexing, and large-scale competitive research. **Caveats: reasoning mode permanently locked to "max" — cannot be reduced, inflating cost and latency on every query. Very slow (28 t/s, 4.02s TTFT). $15.00/M output is the most expensive in the Moonshot lineup (3.75× K2.6). Chatbot Arena early evals (codename "Kivine") showed Claude Fable 5 outperforming on robust UI execution and high-difficulty logic. Inherits K2.6's agentic safety lineage (17/20 on Anthropic's record-tampering eval). Chinese jurisdiction: real-name auth required, non-personal data licensed perpetually to Moonshot AI and affiliates. SWE-bench Verified, MMLU-Pro, and MATH-500 remain pending.**
Architecture Type: Dense Transformer
Standard dense autoregressive transformer architecture with full-attention mechanisms.
Recommended Workloads & Primary Use Cases
long-horizon agentic coding (SWE Marathon
multi-file refactors)
enterprise automation
1M-context research/synthesis
BrowseComp-class retrieval
financial analysis
and agent-swarm coordination — comprehensive codebase reviews
multi-document financial audits
patent/legal indexing
inflating cost and latency on every query
ARMES Zero Data Retention
Calls to Kimi K3 are routed through strict Zero Data Retention inference channels. Prompts and outputs are never stored, indexed, or monitored by Moonshot AI.
