Ce que DigitalOcean a publié en septembre 2026

Now Generally Available: Isolated Worker Nodes for DigitalOcean Kubernetes

DigitalOcean launched generally available Isolated Worker Nodes for its DigitalOcean Kubernetes (DOKS) service, eliminating public IPv4 addresses on worker nodes. Traffic is routed through a NAT Gatew…

How to Manage Resource Limits for Organization Teams

DigitalOcean introduced a new feature allowing organization owners and admins to view, adjust, and request increases for team resource limits directly from the organization dashboard. The feature incl…

30 September 2026 (postgresql)

DigitalOcean's PostgreSQL Advanced Edition no longer allows installing custom extensions via CREATE EXTENSION, though pre-installed extensions like pg_repack, pg_stat_statements, pgaudit, plpgsql, and…

30 September 2026 (organizations)

DigitalOcean introduced a new Resource limits tab in the organization dashboard, allowing owners and admins to view and adjust per-resource usage and limits for each team. This change centralizes reso…

30 September 2026 (inference)

DigitalOcean now supports OpenAI's GPT-6.1 Sol model on its Inference platform, enabling serverless inference and Agent Development Kit use cases. The model is available immediately for customers leve…

30 September 2026 (accounts)

DigitalOcean introduced a new Session duration setting allowing users to configure how long their account sessions remain active, ranging from 1 hour to 30 days. The setting is accessible via the My A…

PostgreSQL and MySQL Plan Changes Beginning 15 October 2026

DigitalOcean will restructure its PostgreSQL and MySQL database plans on 15 October 2026, consolidating larger, high-availability, and read-only options into the Advanced Edition. Standard Edition wil…

30 September 2026 (postgresql, mysql, upcoming)

DigitalOcean will restrict PostgreSQL and MySQL Standard Edition plans to a maximum of 4 GiB RAM starting 15 October 2026 for new accounts and 30 November 2026 for existing accounts. Plans exceeding t…

29 September 2026 (kubernetes)

DigitalOcean Kubernetes (DOKS) now offers isolated worker nodes in general availability, removing all worker nodes from the public internet by default. This enhances security by ensuring nodes are not…

Now Available: Claude Sonnet 5.5 from Anthropic

DigitalOcean now offers Anthropic’s Claude Sonnet 5.5 via its Serverless Inference platform at the same per-token pricing as Sonnet 5. The upgrade improves coding, writing, and document generation cap…

How DigitalOcean Manages Credentials for Autonomous Agents

DigitalOcean introduced a credential management system for its Managed Agents platform that prevents autonomous agents from ever holding raw secrets. Credentials are brokered at execution time via the…

28 September 2026 (valkey)

DigitalOcean now defaults to Valkey 9 for new database clusters, while Valkey 8 remains available for both new and existing clusters. The change applies to managed Valkey services without disrupting e…

28 September 2026 (inference)

DigitalOcean now supports Anthropic’s Claude Sonnet 5.5 model in its Inference service, enabling serverless inference and Agent Development Kit use cases. The model is available immediately for custom…

Now Available: NVIDIA Nemotron 3 Diarization

DigitalOcean now offers NVIDIA Nemotron 3 Diarization, an open-weight streaming speaker diarization model, through its Serverless Inference platform. The model supports real-time speaker attribution f…

DigitalOcean vs Together AI for AI Inference in 2026

DigitalOcean published a detailed comparison of its AI-Native Cloud inference capabilities against Together AI, emphasizing its unified stack (open and frontier models, Inference Router, managed servi…

The agent-first cloud: why we built Managed Agents

DigitalOcean introduced Managed Agents, a new cloud runtime designed for AI-native workloads, combining sandbox, inference, and data into a single platform. The service uses microVMs for bursty comput…

Now in Public Preview: VPC Subnets and Routes

DigitalOcean introduced VPC Subnets and Routes in public preview, enabling network-level isolation and custom routing within VPCs. The feature is available via API today with Control Panel UI support …

25 September 2026 (inference)

DigitalOcean now supports NVIDIA's Nemotron 3 Diarization model on its Inference platform, enabling serverless inference and Agent Development Kit use cases. The model is listed on the Available Model…

What is Jev (2026)? TypeSafe AI's System One model

DigitalOcean added TypeSafe AI’s Jev model to its Model Catalog via Serverless Inference, enabling fast, typed decision-making for classification, routing, and moderation tasks. Jev responds in 70–500…

Now Available: Jev from TypeSafe AI

DigitalOcean now offers Jev, TypeSafe AI's first 'System One' model, through its Serverless Inference service. Jev provides typed decisions with calibrated probabilities and schema conformance, priced…

Best Clouds for AI Model Deployment in 2026

DigitalOcean’s official guide positions its AI-Native Cloud as a top choice for AI model deployment, emphasizing its Inference Router, 70+ supported models, and OpenAI-compatible API. It highlights se…

Now Available: GPT-6 Sol and Luna from OpenAI

DigitalOcean now offers OpenAI’s GPT-6 Sol and Luna models via its Serverless Inference platform. Sol targets complex coding and agentic workflows at $2/M input and $10/M output, while Luna serves hig…

Now Available: Claude Opus 5.5 from Anthropic

DigitalOcean now offers Anthropic’s Claude Opus 5.5 via its Serverless Inference platform, replacing Opus 5 with significant improvements in agentic coding, long-running tasks, and knowledge work. The…

Harness Runtime is now in Public Preview

DigitalOcean launched Harness Runtime in public preview, running each agent session in an isolated Firecracker microVM with built-in pause, resume, checkpoint, and fork capabilities. Users can deploy …

Action Gateway is now in Public Preview

DigitalOcean launched Action Gateway in Public Preview, providing a single managed MCP endpoint for agents to securely access over 16,000 tools. Credentials are brokered at execution time and never ex…

Introducing DigitalOcean Managed Agents: One AI-native stack to power your intelligence

DigitalOcean introduced Managed Agents, a public preview service enabling teams to deploy AI agents (e.g., OpenCode, Codex CLI) with 16,000+ tool integrations, native Inference Engine integration, and…

News Details

DigitalOcean introduced Managed Agents, a new service providing isolated harness runtimes, governed tool access to 16,000+ tools, and Serverless Inference across 75+ models. The platform bills only fo…

22 September 2026 (inference)

DigitalOcean now supports OpenAI’s GPT-6 Sol and Luna models on its Inference platform for serverless inference. The models are immediately available to users via DigitalOcean’s Inference service.

DigitalOcean now supports Anthropic’s Claude Opus 5.5 model in its Inference service, enabling serverless inference, Agent Development Kit, and agent deployments. The model is available immediately fo

22 September 2026 (networking)

DigitalOcean announced VPC subnets and routes are now in public preview, enabling users to add IP ranges to existing VPC networks and direct subnet traffic to managed targets. This expands VPC functio…

What Is an Agent Harness? Architecture and Setup in 2026

DigitalOcean introduced Harness Runtime, a managed execution environment for AI agents that persists state between steps and provides execution history logs, addressing production reliability issues. …

21 September 2026 (managed-agents, agent-harness-runtime, action-gateway)

DigitalOcean introduced Managed Agents in public preview, providing each agent with a dedicated microVM execution environment and governed tool access to scale agentic workloads. The feature aims to e…

Now Generally Available: Managed Databases Advanced Edition for PostgreSQL and MySQL

DigitalOcean announced the general availability of Managed Databases Advanced Edition for PostgreSQL and MySQL, featuring automated proxy-based failover, storage scaling up to 32 TB, and 46% lower sto…

The Next Step for Mission-Critical Workloads: Managed Databases Advanced Edition

DigitalOcean announced general availability of Managed Databases Advanced Edition for MySQL and PostgreSQL, addressing high-growth workloads with up to 38% higher throughput and 50% lower p99 latency.…

17 September 2026 (postgresql, mysql)

DigitalOcean announced the general availability of PostgreSQL Advanced Edition and MySQL Advanced Edition managed database clusters on September 17, 2026. These advanced editions provide enhanced feat…

Now Available: DeepSeek-V4.1-Flash

DigitalOcean now offers DeepSeek-V4.1-Flash via its Inference Engine, a lightweight model using a Causal Encoder-Decoder architecture with just 8B input and 16B output parameters. It reduces KV cache …

What Is SGLang? 2026 Guide to the LLM Serving Framework

DigitalOcean published a guide on SGLang, an open-source LLM serving framework optimized for multi-step agentic workloads, featuring RadixAttention for KV cache reuse. The post emphasizes DigitalOcean…

9 Best LLM API Providers for Developers in 2026

DigitalOcean published a 2026 guide comparing LLM API providers, emphasizing its own Inference Engine with a single OpenAI/Anthropic-compatible endpoint, 70+ models, and an Inference Router for task-a…

Intelligence is yours. Let's keep it that way.

DigitalOcean CEO Paddy Srinivasan argues that closed AI systems risk concentrating intelligence in a few companies and calls for an 'Open Intelligence' movement. The company commits to open-source AI …

DigitalOcean Kubernetes (DOKS) now supports Spot GPU Droplets

DigitalOcean added Spot GPU Droplets support to its managed Kubernetes service (DOKS), enabling automatic interruption handling via cordoning, draining, and PodDisruptionBudget enforcement. Cluster au…

DigitalOcean vs OpenRouter in 2026: AI Routing Comparison

DigitalOcean’s AI-Native Cloud now includes an Inference Router that resolves task intent directly, hosting most open models on its own infrastructure with automatic failover, while OpenRouter acts as…

Built for agents: Omarchy's pipeline moves to DigitalOcean

DigitalOcean now runs Omarchy’s production pipeline and has joined the Omacom Foundation as a Founding Corporate Patron, serving as Omarchy’s agentic compute provider. A 1-click Omarchy Droplet for Di…

Now Available: GPT-6 Astra

DigitalOcean now offers OpenAI's GPT-6 Astra model through its Inference Engine, claiming the largest gains in agentic computer use and coding while reducing output tokens per task. The model is acces…

DigitalOcean vs Fireworks AI for AI Inference in 2026

DigitalOcean positions its AI-Native Cloud as a full-stack alternative to Fireworks AI’s serverless inference platform, emphasizing integrated billing, zero data retention, and built-in model routing.…

Now Available: Memphis Data Center

DigitalOcean introduced its Memphis (MEM1) data center, expanding its global footprint to 21 data centers across 11 regions. MEM1 supports all core products, including Droplets, Kubernetes, Volumes, a…

Now Available: Claude Fable 5.1

DigitalOcean now offers Claude Fable 5.1 across its Serverless Inference, Agent Platform, and ADK, delivering a 13.8-point jump on Terminal-Bench 4.0 and a 75% reduction in cache-read pricing. Anthrop…