November 2026 Summaries

DigitalOcean announced that GPT-5 is available through its serverless inference APIs, allowing developers to use the model without managing GPUs or infrastructure, or to integrate it using their own OpenAI API key. The company positions GPT-5 as a model with improved reasoning, specialization, and performance for applications including financial planning, medical document analysis, code generation, creative work, copilots, research tools, and generative experiences. Existing customers can access the model through serverless APIs, the Gradient AI Platform SDK, or Gradient AI agent workflows, while new users must create a DigitalOcean account and generate a model access key.

DigitalOcean has added Anthropic Claude Opus 4.6 to its Serverless Inference service, allowing developers to access the model through its API, Cloud Console, and Model Playground without managing underlying infrastructure. The model offers a one-million-token context window, adaptive reasoning, and agentic coding capabilities intended for tasks such as large-scale codebase analysis, research, financial and document analysis, multi-agent automation, information retrieval, and generating reports, spreadsheets, and presentations. DigitalOcean positions the service as an integrated option for customers already using its applications, networking, storage, Kubernetes, managed databases, and App Platform, with unified billing, managed inference, and security-hardened project defaults.

DigitalOcean has introduced langchain-gradient, an open-source package that integrates LangChain’s large language model orchestration tools with DigitalOcean’s AI infrastructure. The integration enables developers to use DigitalOcean as a provider in LangChain pipelines, access multiple GPU-accelerated serverless inference models, and retain compatibility with current LangChain versions. Intended uses include retrieval-augmented chatbots, automated customer support, developer copilots, and experimentation with LLM tasks such as summarization, classification, and content generation. The package is available through PyPI, with DigitalOcean planning additional features to improve flexibility and performance for advanced AI workflows.

DigitalOcean has introduced native integrations with LlamaIndex to simplify development of retrieval-augmented generation applications using its managed Knowledge Base and Gradient AI Platform services. Two PyPI packages enable developers to use DigitalOcean Knowledge Bases as LlamaIndex retrievers, with hybrid keyword and semantic search, metadata filtering, and asynchronous operations, and to access Gradient-hosted LLMs with streaming and async support. The integrations shift tasks such as document ingestion, chunking, embeddings, vector storage, and infrastructure management to DigitalOcean, allowing users to add managed retrieval and language models to existing LlamaIndex query or chat workflows with minimal setup. Suggested use cases include documentation-based support assistants, internal knowledge tools, private-repository code assistants, and document research applications.

DigitalOcean has added four fal multimodal AI models to its Serverless Inference platform in public preview, enabling developers to generate images and audio through an API without managing infrastructure, scaling, or vendor integrations. The available models include Stable Diffusion XL Fast and FLUX.1 Schnell for image generation, Stable Audio for text-to-audio creation, and ElevenLabs Multilingual v2 for text-to-speech. Users must opt in through the DigitalOcean console, with access typically granted within 10 to 15 minutes, then submit asynchronous API requests using a model access key. Requests return an ID that can be polled through a status endpoint until complete, after which the generated image or audio result can be retrieved from a separate endpoint.

DigitalOcean has introduced a one-click Marketplace deployment for QM, Y Combinator’s open-source multiplayer agent harness for Slack and web-based workflows, which YC uses across functions including accounting, legal, and engineering. The deployment bundles QM’s core service on a Droplet with Managed Agents sandboxing, Serverless Inference, Managed PostgreSQL for sessions and job queues, and Spaces for files and snapshots under one platform and bill. QM remains designed to let teams select their preferred agent harnesses and models, such as Pi, OpenCode, Codex, or Claude Code, while operating on infrastructure they control. DigitalOcean has also contributed Managed Agents as a built-in sandbox option for supported QM versions, though the service is currently in public preview and is not guaranteed for production workloads.

DigitalOcean has introduced Cuttlefish, a native iPhone app for its Managed Agents platform that lets users monitor, guide, and approve cloud-based AI agent work remotely. The app provides live session streaming, notifications for approvals or failures, detailed reviews of commands and file changes, Face ID or passcode protection for high-risk actions, and controls to pause, resume, stop, or redirect agents. It supports multiple agent frameworks and tools, including Claude Code, Codex CLI, OpenCode, OpenHands, Hermes, LangGraph, and CrewAI, while Managed Agents run in persistent, isolated DigitalOcean microVMs that can be accessed or taken over by teammates. Cuttlefish also integrates with Box Mount, currently in private preview, enabling agents to work directly with synced Box folders while retaining Box permissions, version history, audit trails, and bidirectional file synchronization. DigitalOcean positions the app around choice, control, and collaboration, allowing users to manage long-running agent tasks without needing to remain at their computers.