HOST DIAGNOSTIC · Other
Kimi K2.5
Kimi K2.5· Other· OPEN WEIGHTS
ACTIVE
HOST IDENTIFICATION
Other
Kimi K2.5
ACTIVE
AVG SCORE
[10.8]
CAPABLE
STRONGEST
Cost Efficiency
[13]
WEAKEST
Multimodal Fluency
[9]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
11
Reasoning
11
Mathematical Precision
11
World Knowledge
11
Scientific Acumen
11
Technical
Code Generation
12
Tool Use
11
Multimodal Fluency
9
Speed
11
Cost Efficiency
13
Behavioral
Candor
10
Creativity
10
Tenacity
11
Self-Correction
10
Calibration
10
Operational
Instruction Following
11
Context Fidelity
11
Multi-turn Coherence
11
Planning and Execution
12
Safety Alignment
9
SCORING METHODOLOGY
CALIBRATED. Moonshot AI (China). SWE-bench Verified: 76.8% → Code Gen 16 (75-89% band). Agent Swarm technology: coordinates up to 100 sub-agents in parallel, reducing execution time for complex coding by up to 4.5x. On Arena Document leaderboard (added April 14). Planning+Execution scored high (16) for multi-agent coordination. Kimi K2.5 Thinking variant available. Cost Efficiency 17 (open weights, competitive pricing). [2026-06-13 CORRECTION: Release Date fixed from placeholder 2026-01-01 → 2026-01-27, the confirmed official launch (sources: Baidu Baike, AWS Bedrock model card, WaveSpeed, kimi.com). Family/Provider remain 'Other' — no Kimi/Moonshot enum option, consistent with Kimi K2 / K2.6 records.]