Cloudflare Workers AI Adds Kimi K2.5 for End-to-End Agent Workflows

Cloudflare

Mar 20, 2026 · Updated Jun 5, 2026

Cloudflare's Workers AI now supports Kimi K2.5, Moonshot AI's frontier open-source model with a 256k context window. Developers can build and run full agent workflows on Cloudflare's platform, with prefix caching and a new async API cutting inference costs.

Cloudflare Workers AI now runs Kimi K2.5, Moonshot AI's frontier open-source model with a 256k context window, multi-turn tool calling, vision inputs, and structured outputs. Workers AI previously focused on smaller models — this marks their entry into large-model inference. Cloudflare built custom kernels on their Infire inference engine, handling the tensor and parallelization optimizations you'd otherwise need a dedicated ML team to manage.

Cloudflare's internal security review agent, processing over 7 billion tokens per day, ran 77% cheaper after switching to Kimi K2.5 compared to a mid-tier proprietary model. As agentic workloads scale, open-source frontier models with this price-performance ratio become the practical path for enterprises running high-volume inference.

The model is available as @cf/moonshotai/kimi-k2.5 on Workers AI. A new x-session-affinity header improves prefix cache hit rates for multi-turn sessions, and a revamped async API handles batch inference without capacity errors.

View the full update on blog.cloudflare.com

Cloudflare

@CloudflareMar 20

Kimi K2.5 is now available on #WorkersAI. You can now build and run agents end-to-end on the Cloudflare Developer Platform. Read about how we tuned our inference stack to drive down costs for internal agent workflows. https://t.co/kEQ6HHpoJS

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Cloudflare →

Keep reading

Fireworks AI Adds Kimi K2.6 Training to Build Custom Frontier Agents

Fireworks AI added Moonshot AI's Kimi K2.6 to its training platform, enabling supervised fine-tuning and reinforcement learning on the 1-trillion parameter model. This allows teams to customize the leading open-weight agentic model for specific production workflows while maintaining a 265K context window.

Moonshot AI Launches Kimi K2.6 with 4000 Step Long Horizon Coding

KimiApr 24

Moonshot AI Launches Kimi K2.6 with 4000 Step Long Horizon Coding

Moonshot AI released Kimi K2.6, an open-source model that achieves state-of-the-art scores on SWE-Bench Pro and Toolathlon. The update introduces long-horizon coding, enabling agents to execute over 4,000 autonomous steps without losing context or drifting from the task.

Cloudflare Integrates GPT-5.5 to Power Persistent Autonomous Agents

CloudflareApr 24

Cloudflare Integrates GPT-5.5 to Power Persistent Autonomous Agents

Cloudflare added OpenAI's GPT-5.5 to its AI Gateway, featuring a 1M token context window and 2x cost efficiency over competing frontier coding models. The model is optimized for agentic loops, enabling systems to plan, use tools, and self-verify their work until a task is complete.

Cloudflare adds MiniMax M3 with 1M context for agentic coding

MiniMaxJun 2

Cloudflare adds MiniMax M3 with 1M context for agentic coding

Cloudflare has integrated the MiniMax M3 foundation model into its AI Gateway platform. The update provides developers with a high-context, multimodal model specialized for autonomous coding tasks directly within their existing infrastructure.