kimi-k2.5

Kimi K2.5 is a native multimodal agentic model designed to advance general agentic intelligence by integrating vision and language understanding. Built upon the Kimi K2-Base, it utilizes a trillion-parameter mixture-of-experts transformer architecture. The model is engineered to jointly optimize text and vision, ensuring these modalities enhance one another rather than competing for capacity. Its primary innovation is the Agent Swarm framework, which allows the model to dynamically decompose complex tasks into smaller, heterogeneous sub-problems and execute them concurrently, significantly improving efficiency over traditional single-agent approaches.

The model underwent continual pre-training on approximately 15 trillion mixed visual and text tokens to refine its reasoning and coding performance. By coordinating up to 100 specialized AI agents simultaneously, the system achieves a 4.5x reduction in execution latency compared to standard baselines. This parallel orchestration makes it well-suited for complex workflows requiring high-level mathematical reasoning, computer use, and multimodal analysis. As an open-weight model, it is intended to support ongoing research and the development of practical, real-world agentic systems.

Ollama Cloudkimi-k2.5kimi-k2

Open provider page Provider docs

Quick Info

Provider: Ollama Cloud
Model key: kimi-k2.5
Release date: Jan 27, 2026
Last updated: Jan 27, 2026
Input modalities
Output modalities
Capabilities

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Latest news about kimi-k2.5

Tencent Coding Plan (China)

CoverageBenchmark

A Lorphic explainer dated July 13, 2026 walks through the Kimi K2 model family (K2, K2.5, K2.6, K2.7), emphasizing that the variants are architecturally related but carry different capabilities, licensing implications, and recommended use cases. It attributes to Moonshot's technical documentation and Hugging Face model The same explainer states the K2 base model was pre-trained on 15.5 trillion tokens and that Moonshot used the Muon optimizer for the family, framing these as the architectural foundation that all K2 variants (including K2.5) share. It positions the article as a practical breakdown covering architecture, version-by-ver

Ollama Cloud

Coverage

Kimi K2.5 brings advanced agentic reasoning, design-to-code capabilities, and production-ready performance to fully managed inference endpoints on Nscale. Follow our blog for more insights from Nscale.

Ollama Cloud

CoverageRelease Notes

Moonshot AI releases open-source Kimi K2.5 model with 1T parameters - SiliconANGLE

Ollama Cloud

Coverage

The open-source AI landscape just received a seismic jolt with the release of Kimi K2.5, an advanced visual agentic intelligence model. AI analyst Matthew Berma

Ollama Cloud

Coverage

Moonshot AI introduces Kimi K2.5, boosting multimodal capabilities and coding performance, as China's AI sector intensifies.

Ollama Cloud

Coverage

Moonshot's Kimi team has officially introduced Kimi K2.5, described as the most powerful open-source model to date and built as a native multimodal system through continued pretraining over approximately 15 trillion mixed visual and text tokens. The model delivers state-of-the-art coding and vision capabilities and int Kimi K2.5 is available via Kimi.ai, the Kimi App, the API, and Kimi Code, with four supported modes on Kimi.ai and the Kimi App: K2.5 Instant, K2.5 Thinking, K2.5 Agent, and K2.5 Agent Swarm (Beta), where Agent Swarm is currently in beta on Kimi.ai with free credits for high-tier paid users. Reported benchmark coverage