Kimi K2.5
Kimi K2.5 by Moonshot AI is a native multimodal agentic model with 1T-param MoE, Agent Swarm execution, vision-grounded coding, and 256K context.
Model Overview
Capabilities, design details, and architectural traits
Kimi K2.5 - Native Multimodal Agentic Model
Kimi K2.5 is an open-source model from Moonshot AI, built via continual pretraining on roughly 15 trillion mixed visual and text tokens atop the Kimi K2-Base checkpoint. Vision-language understanding is trained in from the start, paired with agentic execution and selectable instant/thinking response modes.
Defining Characteristics
- Native Multimodality: vision-language tokens are part of pretraining itself, supporting visual knowledge, cross-modal reasoning, and tool use grounded in images and video.
- Coding with Vision: generates code directly from visual specifications such as UI designs and video workflows.
- Agent Swarm: moves from single-agent execution to a coordinated swarm scheme, decomposing tasks into parallel sub-tasks run by dynamically instantiated, domain-specific agents.
Why It Exists
Kimi K2.5 extends the Kimi K2 architecture into one model handling vision, language, and long-horizon agentic execution together, rather than layering multimodality onto a text-only base.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Kimi | Anthropic | Anthropic |
| Release Date | January 27, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 262K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.60 Best Input Pricing | $4 | $2 |
| Output Pricing | $3 Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | textimage | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 19.4 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | - | - | - |
| Agentic Index | - | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from Kimi
Other models by Kimi
Top AI Models
Leading alternatives by intelligence score
