KIMI K2.5
Open SourceMoonshot AI•Released on 2026-01-27
Moonshot AI's flagship agentic model with native multimodal architecture. Unifies vision and text, thinking and non-thinking modes, single-agent and multi-agent execution. Features visual coding (UI screenshots to code) and self-directed agent swarm paradigm. #2 on Artificial Analysis Intelligence Index among open models.
78
Overall Score
Core Specs
256K
Context Window
32K
Max Output
✓
Reasoning
✓
Open Source
Multimodal Support
textimagevideo
Scenario Scores
User Feedback Highlights
Based on community feedback. Hover to see original reviews.
+ Visual coding capability+ Open weights (Modified MIT)− Very verbose output+ Native multimodal (text, image, video)+ Multi-agent swarm support− Higher cost than K2+ 256K context window− Slower than average (42 tok/s)− Expensive for open model
Sentiment:👍 68%😐 22%👎 10%
Pros & Cons
Pros
- +Native multimodal (text, image, video)
- +Visual coding capability
- +Multi-agent swarm support
- +256K context window
- +Open weights (Modified MIT)
Cons
- −Very verbose output
- −Slower than average (42 tok/s)
- −Higher cost than K2
- −Expensive for open model
Reliability
SLA99.0%
Incidents (30d)1
Last Incident2026-03-15
Supports self-hosting. API servers primarily in China. Available via OpenRouter globally.
View Status Page →Pricing
Input (per 1M tokens)$0.60
Output (per 1M tokens)$3.00
Free trial available
Updated on 2026-04-03
Get Started
1Visit the provider's website
2Create an account
3Start using the model
Compare with Others
Benchmarks
intelligenceIndex47%
sweBench72.5%
terminalBench68.2%
gpqaDiamond71.8%
hle41.2%