Moonshot AI 的 Kimi K3 登陆 DigitalOcean 推理服务

# Moonshot AI's Kimi K3 Launches on DigitalOcean Serverless Inference ## Combined views 90.9K 2 posts, first seen 5d ago # Moonshot AI's Kimi K3 Launches on DigitalOcean Serverless Inference ## Reactions from ranked influencers Kimi K3 is now available on@digitalocean's Serverless Inference! Developers can start building with our most capable model in minutes. .@Kimi\_Moonshot K3 from Moonshot AI is now live on DigitalOcean Inference Engine. 1M-token context, native vision, built to run agentic tasks for hours. Supported on Inference Router and model synthesis to maximize intelligence per dollar. No setup, no model ops. Kimi K3 is now available on @digitalocean 's Serverless Inference! Developers can start building with our most capable model in minutes. [...] ## Reactions from the X community @Kimi\_Moonshot @digitalocean Great to see Kimi K3 becoming easier to access. Lowering the barrier for developers usually leads to more experimentation and better applications. @Kimi\_M

Copy page as Markdown View page as Markdown DigitalOcean Inference supports more than 70 foundation, embeddings, and reranking models, including OpenAI (GPT-5.x and open-weight gpt-oss), Anthropic Claude, Meta Llama, DeepSeek, Alibaba Qwen, Moonshot AI Kimi, NVIDIA Nemotron, and Z.ai GLM. All text models are served through OpenAI-compatible endpoints, so you can use an existing OpenAI SDK or client by changing the base URL to ` and authenticating with a DigitalOcean API key. For endpoint details, see Serverless Inference Endpoints. Note For pricing information, see the pricing page. [...] ### DeepSeek Models on DigitalOcean DigitalOcean hosts DeepSeek V4 Pro and V4 Flash (with input context windows of up to 1M tokens), DeepSeek V3.2, DeepSeek V3, and DeepSeek R1 Distill Llama 70B. ### Kimi K2.5 and K2.6 on DigitalOcean DigitalOcean hosts Moonshot AI’s Kimi K2.5 and Kimi K2.6 for serverless and dedicated inference, with prompt caching and support for the Chat Completions and Re

DigitalOcean logo DigitalOcean ## 27 July DigitalOcean adds Moonshot AI’s Kimi K3 model to Inference for serverless, dedicated inference, ADK, and agents. ### The following Moonshot AI model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents: + Kimi K3 For more information, see the Available Models page. Original source All of your release notes in one feed Join Releasebot and get updates from DigitalOcean and hundreds of other software products. Create account Get updates with: Jul 24, 2026 + Date parsed from source: Jul 24, 2026 + First seen by Releasebot: Jul 25, 2026 DigitalOcean logo DigitalOcean ## Now Available: Claude Opus 5 from Anthropic [...] Releasebot Docs Pricing Browse feeds Latest Sign in Create account # DigitalOcean Release Notes Follow Follow DigitalOcean to add their release notes to your feed! 154 release notes curated from 16 source

## What’s next for Kimi K3 on DigitalOcean Kimi K3 is available today on DigitalOcean Inference Engine. Access it through Serverless Inference for fully managed, usage-based inference with no infrastructure to operate, or add it to your existing model mix with the Inference Router to intelligently route requests based on cost, latency, or workload. We’re excited to see what customers build with K3, and we’ll be sharing more soon about how teams are using it in production. Want to see Kimi K3 in action? We built the same AI agent with and without Kimi K3, then compared the results side by side. See what changed, what didn’t, and where Kimi K3 made the biggest difference. Check it out: We Built the Same App Twice With and Without Kimi K3 [...] Standing up a new model, integrating it into DigitalOcean’s Inference Engine, and showcasing its unique attributes on day 0 takes three things: the right hardware, a tuned serving stack, and rigorous verification against Moonshot’s own benchmarks

You can now access DigitalOcean Serverless Endpoints directly from build.nvidia.com, letting you experiment with elite open‑weight models like Zhipu AI GLM‑5, Moonshot AI Kimi‑K2.5, and MiniMax‑M2.5, and other frontier models from DeepSeek, Meta, and OpenAI. Start building with NVIDIA NemoClaw and the Agent Toolkit on build.nvidia.com, then deploy seamlessly to DigitaOcean’s fully managed Serverless Inference, eliminating infrastructure headaches and giving you more time to focus on shipping AI-powered products. Product Update Now Available: Open-Weight Models on DigitalOcean Serverless Inference You can now run these powerful open-weight models directly on DigitalOcean Serverless Inference, fully integrated with your existing environment: [...] Models available now: DeepSeek-V3.2 DeepSeek-V4-Pro DeepSeek-V4-Flash Kimi K2.6 Kimi K2.5 GLM-5 GLM-5.1 GLM-5.2 gpt-oss-120b Qwen 3.5 MiMo-V2.5 MiMo-V2.5-Pro MiniMax M2.5 Qwen 3 Coder Read the documentation to learn how to str

If you choose to go with serverless inference, DigitalOcean offers Kimi K3 on the Serverless Inference API as a pay-per-token option at the $3/$15 rate. If you choose to go the self-hosted GPU route, you can use DigitalOcean’s GPU Droplets, Dedicated Inference, or long-term reserved GPUs. With GPU Droplets, you manage the stack, and with Dedicated Inference, DigitalOcean runs a managed vLLM endpoint that you bring the Kimi K3 weights to. The 4-bit native options DigitalOcean provides are the AMD Instinct MI350X and the NVIDIA B300. Availability can change without notice, so be sure to check what is currently available and on which regions before planning your deployment. [...] Kimi K3’s first-party API is $3.00 input / $0.30 cached input / $15.00 output per 1M tokens, with DigitalOcean live at the same $3/$15. The cheapest native-FP4 single node is ~8x MI350X at ~$4.76/GPU-hour. This is approximately $38/hr and $27,800/month on a reserved GPU Droplet that is running whether or not anyo