DigitalOcean Docs

Private Droplets are now generally available in all regions. Private Droplets have no public network interface and no public IP address, using VPC-only networking with automatic integration with VPC NAT gateway, VPC peering, and VPC private DNS. See the Private Droplets documentation for setup instructions and limitations.

The following agent evaluation metrics are deprecated and should no longer be used:

• Tone
• Retrieved Chunk Usage
• Prompt Perplexity

Use the currently supported metrics listed in Agent Evaluation Metrics instead. To monitor deployed agent behavior outside of evaluations, use Agent Metrics and Runtime Logs.

The following Anthropic model is now available on DigitalOcean Inference for serverless inference, Agent Development Kit, and agents:

• Claude Sonnet 5

For more information, see the Available Models page.

Agent evaluations support for the Agent Development Kit (ADK), previously in preview, is now removed.

To evaluate agents, use agent evaluations via the DigitalOcean Control Panel for supported agent types. To monitor ADK agent behavior, use Agent Metrics and Runtime Logs.

Insights, agent tracing, and conversation logs are deprecated for all agents, including agents created through the Control Panel, CLI, API, and Agent Development Kit (ADK).

To monitor deployed agent behavior, use Agent Metrics and Runtime Logs instead.

The Agent Evaluations MCP server tool has been renamed to Evaluations.

Custom metrics are now available for DigitalOcean Evaluations. You can define your own metrics to evaluate model behavior against criteria specific to your use case.

DigitalOcean Evaluations is now generally available. Use Evaluations to create test cases, run evaluation datasets, and measure model performance against selected metrics.

Presets are now available for DigitalOcean Evaluations. You can save and reuse evaluation configurations, including the candidate model, system prompt, hyperparameters, judge model, and metrics.

Model Evaluations is now renamed to DigitalOcean Evaluations.

Serverless Inference now requires a positive prepaid account balance before you can send inference requests. Usage charges are deducted from this balance, and access is suspended if it reaches $0. You can add a prepayment manually or enable auto-reload to replenish your balance automatically. For more information, see Manage Serverless Inference Prepayment.

The DigitalOcean Control Panel now supports light and dark themes. From the profile menu in the top right corner of the control panel, you can set your theme to a light or dark appearance, or match your operating system’s appearance setting.

You can now attach Network File Storage shares to multiple VPC networks, and expose specific subdirectories to single VPC networks by creating access points.

A share can connect to up to 10 VPC networks in the same region. Access points restrict clients to a specific path within the share, and are isolated from one another so that clients on one cannot see the directories governed by another. This lets you host multiple tenants on a single share, each scoped to their own directory. For details, see How to Create and Delete Network File Storage Access Points.

Single sign-on (SSO) with OIDC for DigitalOcean Kubernetes (DOKS) clusters is now in general availability. You can authenticate users to your Kubernetes clusters through an identity provider like Auth0, authentik, JumpCloud, Keycloak, or Okta, instead of using token-based authentication.

SSO is configured per cluster with an issuer URL and client ID from your identity provider. You can enable it using doctl, the DigitalOcean API, or Terraform.

The following Z.ai model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• GLM-5.1

For more information, see the Available Models page.

A Fedora 44 (fedora-44-x64) Droplet base image is now available in the Control Panel and through the API.

The following Z.ai model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• GLM-5.2

For more information, see the Available Models page.

To comply with Peru’s tax regulation for digital services, DigitalOcean charges Value Added Tax (VAT) for customers with a tax location in Peru, beginning on 1 June 2026. These charges appear on invoices issued on and after 1 July 2026. Learn more about taxes in Peru.

The following Xiaomi model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• MiMo-V2.5-Pro

For more information, see the Available Models page.

DigitalOcean Inference supports server-side tools on serverless inference, dedicated inference, and inference routers. You can add the following tools:

• Web search, web fetch, knowledge base retrieval, and remote MCP server tools to your requests in the Chat Completions and Responses APIs.
• Provider-native tools such as bash, text editor, computer use, and web fetch for Anthropic models with the Messages API.
• Function calling and tool search for OpenAI models on the Responses API, and Anthropic models on the Messages API.

Web search and web fetch are in public preview. For more information, see Use Server-Side Tools.

Updated CentOS Stream 9 and CentOS Stream 10 (centos-stream-9-x64, centos-stream-10-x64) Droplet base images require a minimum Droplet size of s-1vcpu-1gb. You cannot create Droplets with these images on the s-1vcpu-512mb-10gb plan because the image exceeds the available disk size for that plan. For more information, see Linux Images for Droplets.

The following Anthropic models are deprecated from DigitalOcean Inference as of 15 June 2026:

• Claude Opus 4
• Claude Sonnet 4

Migrate to Claude Opus 4.8 (anthropic-claude-opus-4.8) and Claude Sonnet 4.6 (anthropic-claude-4.6-sonnet), respectively, to avoid service disruption. For information on our model deprecation policy and recommended replacement models, see Model Support Policy.

The following Xiaomi model is now available on DigitalOcean Inference for serverless inference, Agent Development Kit, and agents:

• MiMo-V2.5

For more information, see the Available Models page.

Claude Fable 5 is no longer available on DigitalOcean Inference. Access to all other Anthropic models remains available. For more information, see the Anthropic statement on Claude Fable 5.

We support passthrough tool search on the Messages API for Anthropic models and the Responses API for OpenAI models, enabling deferred loading of tools in agentic workflows. There is no additional cost to using tool search. For more information, see Use Server-Side Tools.

Debian 12 reached end of life on 10 June 2026. Per our image deprecation policy, this image is available exclusively via the API for the next 30 days before we remove it from our platform.

The following Anthropic model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• Claude Fable 5

For more information, see the Available Models page.

The following NVIDIA model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• Nemotron 3 Ultra

For more information, see the Available Models page.

The public endpoint when creating a dedicated inference deployment is now disabled by default. For more information on how to enable access, see Use Dedicated Inference Endpoint.

You can now use the DigitalOcean Control Panel to create, connect to, tag, resize, and restore Managed Weaviate clusters (in private preview). See the Managed Weaviate docs for more information.

The following Anthropic model is now available on DigitalOcean Inference for serverless inference, Agent Development Kit, and agents:

• Claude Opus 4.8

For more information, see the Available Models page.

The following DeepSeek model is now available on DigitalOcean Inference for serverless inference, dedicated inference, Agent Development Kit, and agents:

• DeepSeek V4 Flash

For more information, see the Available Models page.

Fedora 42 has reached end of life. Per our image deprecation policy, this image is available exclusively via the API for the next 30 days before we remove it from our platform.

Request-based autoscaling for App Platform is now in general availability. Service components can scale automatically based on HTTP traffic metrics, including requests per second and P95 request duration, in addition to or instead of CPU utilization. Request-based autoscaling works with both shared and dedicated CPU plans.

Configure request-based autoscaling in the DigitalOcean Control Panel, with the API, or with doctl.

MySQL 8.4 is now available for database clusters. New clusters use MySQL 8.4 by default. For version support and MySQL 8.0 end of life, see MySQL Limits.

DigitalOcean Kubernetes (DOKS) now supports single sign-on (SSO) with OIDC in public preview. You can authenticate users to your Kubernetes clusters through an identity provider like Auth0, JumpCloud, Keycloak, or Okta, instead of using token-based authentication.

SSO is configured per cluster with an issuer URL and client ID from your identity provider. You can enable it using doctl.

Registered businesses in Tanzania may now be required to withhold 15% of the gross amount of payments for digital services from DigitalOcean under the Tanzania Income Tax Act. Learn more about taxes in Tanzania.

The following Moonshot AI model is now available on DigitalOcean Inference for serverless inference, Agent Development Kit and agents:

• Kimi K2.6

For more information, see the Available Models page.

A remote MCP server is now available for Functions, providing API-based access for AI tools to manage DigitalOcean resources.

The following DeepSeek model is now available on DigitalOcean Inference for serverless inference, Agent Development Kit and agents:

• DeepSeek V4 Pro

For more information, see the Available Models page.

NVIDIA B300 GPUs are now available as multi-node GPU worker nodes in DigitalOcean Kubernetes (DOKS), by contract only. To add single- or multi-node B300 GPU worker nodes to your cluster, contact sales. See GPU Worker Nodes for all available GPU types and node pool slugs.

DigitalOcean Vector Databases is now generally available in Data Services that groups managed engines for vector similarity search. The launch includes:

• Weaviate in private preview for retrieval-augmented generation and semantic search workloads. Available to opted-in customers; preview clusters are not billed.
• OpenSearch with the k-NN, ML Commons, and Neural Search plugins for hybrid (vector plus keyword) search and remote embedding models. Uses the existing Managed Databases OpenSearch engine.
• PostgreSQL with the vector (pgvector) and vectorscale (pgvectorscale) extensions for vector similarity search alongside relational data.

For an overview and guidance on choosing an engine, see Vector Databases.

DigitalOcean Managed Weaviate is now in private preview. Opted-in customers can provision Weaviate clusters in the TOR1 region in Small, Medium, and Large plans through the dedicated DigitalOcean Vector Databases API at /v2/vector-databases. Clusters are reachable over port 443 for both HTTP and gRPC and support configurable quantization (rq, pq, bq, or sq) for the vector index.

During preview, cluster management is API-only. Preview clusters are not billed and are not covered by a paid support SLA. APIs, SKUs, regions, and Control Panel elements may change before general availability. For setup and usage guidance, see Managed Weaviate.

A remote MCP server is now available for Volumes Block Storage, providing API-based access for AI tools to create, attach, detach, and manage volumes and volume snapshots.

PostgreSQL Advanced Edition clusters are now available in public preview, offering enhanced performance and scalability for production workloads.

You can now use DigitalOcean personal access tokens for authenticating serverless inference requests. You can use a personal access token as an alternative to a model access key when sending requests to the serverless inference API. Model access keys remain recommended when you need per-application scoping, VPC restriction, or credentials dedicated to inference workloads. For more information, see Serverless Inference Overview.

A remote MCP server is now available for Network File Storage, providing API-based access for AI tools to create and manage NFS shares and access rules.

DigitalOcean AI Agentic Cloud is now DigitalOcean AI-Native Cloud.

MySQL Advanced Edition clusters are now available in public preview, offering enhanced performance and scalability for production workloads.

Launchpad Starter Kits are now generally available. Available kits include RAG Assistant, Data Workflow, and Observability. Each kit deploys a preconfigured application stack with opinionated defaults, so you can get started without manually provisioning or configuring the underlying resources.

The following Beijing Academy of Artificial Intelligence (BAAI) embeddings model is now available in Data Services for DigitalOcean Knowledge Bases:

• BGE M3

For more information, see the Available Models page.

The following Beijing Academy of Artificial Intelligence (BAAI) reranking model is now available in Data Services for DigitalOcean Knowledge Bases:

• BGE Reranker (v2) M3

For more information, see the Available Models page.

The following IntFloat embeddings model is now available in Data Services for DigitalOcean Knowledge Bases:

• E5 Large (multilingual)
• E5 Large (v2)

For more information, see the Available Models page.

RAG Playground is now available in DigitalOcean Knowledge Bases. It lets you run queries against a knowledge base and test how a serverless inference model generates answers from retrieved content.

For more information, see the Data Services Features page.

Data Services now supports reranking for DigitalOcean Knowledge Bases to improve the relevance of retrieved results before they’re returned or used in generated responses. For more information, see Create Knowledge Bases and Test Knowledge Bases.

Data Services now lets you retrieve data from DigitalOcean Knowledge Bases using the Control Panel with semantic, keyword, or hybrid searches, apply filters, review retrieved chunks, and copy live code examples. For more information, see Test Knowledge Bases.

DigitalOcean Knowledge Bases are now in General Availability in Data Services. Using knowledge bases, you can store, index, and retrieve data for AI applications.

DigitalOcean Knowledge Base retrieval is now available through a DigitalOcean MCP server.

The following Beijing Academy of Artificial Intelligence (BAAI) reranking model is now available on DigitalOcean Inference for DigitalOcean Knowledge Bases:

• BGE Reranker (v2) M3

For more information, see the Available Models page.

The following embeddings model is now available on DigitalOcean Inference for DigitalOcean Knowledge Bases:

• E5 Large (multilingual) (IntFloat)
• E5 Large (v2) (IntFloat)
• BGE M3 (Beijing Academy of Artificial Intelligence (BAAI))

For more information, see the Available Models page.

Knowledge base enhancements are now generally available in DigitalOcean Inference, including the updated creation workflow, chunking controls, and data retrieval for testing knowledge base. For more information, see Create and Manage Agent Knowledge Bases.

RAG Playground is now available in DigitalOcean Inference for DigitalOcean Knowledge Bases. It lets you run queries against a knowledge base and test how a serverless inference model generates answers from retrieved content.

For more information, see the DigitalOcean AI Platform Features page.

As part of the DigitalOcean AI-Native Cloud, DigitalOcean Gradient™ AI Platform is now DigitalOcean AI Platform.

DigitalOcean Inference now supports reranking for knowledge bases to improve the relevance of retrieved results before they’re returned or used in generated responses. For more information, see Create and Manage Agent Knowledge Bases and Test Reranking.

DigitalOcean Inference now lets you retrieve data from knowledge bases using the Control Panel with semantic, keyword, or hybrid searches, apply filters, review retrieved chunks, and copy live code examples. For more information, see Create and Manage Knowledge Bases.

Dedicated Inference is now in General Availability.

• A remote MCP server is also available, allowing MCP clients to create, update, list, and delete Dedicated Inference endpoints. For more information, see Dedicated Inference MCP Tools.

You can now browse Model Catalog through a DigitalOcean MCP server.

Batch inference lets you submit text-only batch jobs for OpenAI and Anthropic models. Using batch inference significantly reduces cost compared to real-time inference. For more information, see Use Batch Inference.

Bring Your Own Models (BYOM) is now available in Model Catalog. You can import models from Hugging Face or Spaces buckets or folders. For details, see Import a Model.

Model Catalog is now in General Availability.

You can now evaluate models available for serverless inference, inference routers, and dedicated inference deployments using a judge model. Scoring includes metrics such as correctness, completeness, ground truth faithfulness, and safety metrics. This features is in public preview. You can opt in from the Feature Preview page. For more information, see Evaluate Models.

We now support multimodal models for serverless inference. Multimodal models process and generate content across multiple data types, including images, audio, video, and text, thus enabling a much broader range of real-world applications, including document intelligence, voice agents, content generation, and accessibility tools. For more information, see Use Multimodal Inference.

The Model Playground now supports the following features when testing and comparing models:

• Uploading images from local storage

• Generating multimodal artifacts, such as images, audio, and text-to-speech, from models that support it

Read Test and Compare Models for more information.

The following models are now available on DigitalOcean Inference:

• Qwen3 Coder Flash (Alibaba)
• DeepSeek V3.2 (DeepSeek)
• Gemma 4 (Google)
• Llama 4 Maverick 17B 128E Instruct (Meta)
• Ministral 3 14B Instruct (Mistral AI)
• Nemotron Nano 12B v2 VL (NVIDIA)
• Nemotron Nano 3 Omni (NVIDIA)
• BGE M3 (BAAI)
• E5 Large (multilingual) (Intfloat)
• Qwen 3 TTS (1.7B) (text-to-speech)
• Wan2.2-T2V-A14B (text-to-video)
• Stable Diffusion 3.5 Large (image generation)

For more information, see the Models page.

As part of the DigitalOcean AI-Native Cloud, DigitalOcean AI Inference Hub is now DigitalOcean Inference.

Inference Router in now available in public preview and enabled for all users. Using this feature, you can use multiple models in a model pool to configure routing rules and selection policy for inference requests. We provide pre-built templates or you can define custom task-matching logic using natural language, with configurable fallback support for reliability. For more information, see Inference Router.

DigitalOcean Inference now supports scoped model access keys. When you create a key, you can limit it to specific foundation models and inference routers, enable batch inference, and restrict it to a VPC network so that only requests from that VPC network can authenticate. Team owners can also view and manage keys created by other team members. Previously created keys continue to authenticate without changes. For more information, see Model Access Keys.

As part of the DigitalOcean AI-Native Cloud, DigitalOcean AI GPU Droplets is now GPU Droplets.

Data Services is now generally available, providing managed databases and knowledge bases for storing, indexing, and retrieving application data.

Managed Databases are now generally available in the DigitalOcean Control Panel under Data Services, where you can create and manage the following database engines:

• Kafka
• MongoDB
• MySQL
• OpenSearch
• PostgreSQL
• Valkey

As part of the DigitalOcean AI-Native Cloud, DigitalOcean AI Bare Metal GPUs is now Bare Metal GPUs.

The following OpenAI model is now available on DigitalOcean AI Platform for Agent Development Kit and agents:

• GPT-5.5

For more information, see the Available Models page.

The following OpenAI model is now available on Inference for serverless inference:

• GPT-5.5

For more information, see the Available Models page.

DigitalOcean Container Registry now supports container images up to 100 GB.

DigitalOcean Container Registry now supports image layers up to 20 GB.

A remote MCP server is now available for DigitalOcean Container Registry (DOCR), providing API-based access for AI tools to manage your container registries, repositories, and garbage collection.

Now in public preview, App Platform supports request-based autoscaling for service components. Services can now scale automatically based on HTTP traffic metrics, including requests per second and P95 request duration, in addition to or instead of CPU utilization. Request-based autoscaling works with both shared and dedicated CPU plans.

The following OpenAI model is now available on DigitalOcean Gradient™ AI Platform for serverless inference and Agent Development Kit:

• GPT Image 2

For more information, see the Available Models page.

The following OpenAI model is now available on DigitalOcean AI Inference Hub for serverless inference:

• GPT Image 2

For more information, see the Available Models page.

DigitalOcean Gradient™ AI Platform is now DigitalOcean AI Platform.

The following Anthropic model is now available on DigitalOcean Gradient™ AI Platform for serverless inference, Agent Development Kit, and agents:

• Claude Opus 4.7

For more information, see the Available Models page.

The following Anthropic model is now available on DigitalOcean AI Inference Hub for serverless inference:

• Claude Opus 4.7

For more information, see the Available Models page.

AMD Instinct MI325X GPUs are now available in TOR1 by contract only, in 1- and 8-GPU configurations for single- and multi-node GPU Droplets. To create GPU Droplets and DOKS GPU worker nodes with MI325X GPUs, contact sales. Learn more about GPU Droplet plans and DOKS GPU worker nodes.

You can now transfer Droplet snapshots between accounts using the DigitalOcean API. Read How to Transfer a Droplet Snapshot to learn how to transfer snapshots using the API or Control Panel.

We have updated the following buildpacks for App Platform:

• Go buildpack: We added the following Go versions. If you have an existing Go app that uses v0, we recommend upgrading to v1.
  • go1.24.12 - go1.24.13
  • go1.25.6 - go1.25.8
  • go1.26.0 - go1.26.1
• PHP buildpack: We added the following PHP versions. If you have an existing PHP app that uses v1, we recommend upgrading to v2.
  • PHP 8.3.29 - 8.3.30
  • PHP 8.4.16 - 8.4.17
  • PHP 8.5.0 - 8.5.2
• Python buildpack: We updated supported Python aliases and package tooling. If you are on Ubuntu-22 and have an existing Python app that is on v4, v3, v2, v1, or v0, we recommend upgrading to v5.
  • Python 3.10 alias now resolves to 3.10.20
  • Python 3.11 alias now resolves to 3.11.15
  • Python 3.12 alias now resolves to 3.12.13
  • Python 3.13 alias now resolves to 3.13.12
  • Python 3.14 alias now resolves to 3.14.3
  • Updated uv to 0.10.9
• Node.js buildpack: We added the following Node.js versions. Visit the Node.js buildpack to learn more about specifying a Node.js engine version.
  • Node.js 20.20.0 - 20.20.1
  • Node.js 22.22.0 - 22.22.1
  • Node.js 24.13.0 - 24.14.0
  • Node.js 25.3.0 - 25.8.1
• Ruby buildpack: We added the following Ruby versions. If you have an existing Ruby app that uses v1, we recommend upgrading to v2.
  • Ruby 3.2.10 - 3.2.11
  • Ruby 3.3.11
  • Ruby 3.4.9
  • Ruby 4.0.1 - 4.0.2

NAT gateways now provide 2 Gbps of symmetrical bandwidth for each size, up from 25 Mbps. NAT gateways are available in 1-5 size increments and each size increment includes 100 GiB of outbound data transfer per month.

Read VPC NAT Gateway Features and VPC Pricing for details.

Droplets that use the VPC-local DNS resolver can now access Spaces buckets over DigitalOcean’s internal network instead of the public internet. Traffic routed over the internal network does not count against your Spaces outbound transfer allowance.

Private Spaces traffic is available in all regions; in the AMS, NYC, and SFO region groups, Droplets can also access Spaces buckets in other datacenters within the same group over the internal network. For more details, see the Spaces bandwidth billing documentation.

The following models are deprecated from DigitalOcean Gradient™ AI Platform:

• Meta Llama 3.1 8B-Instruct
• Mistral NeMo

Migrate to a supported active model to avoid service disruption. For information on our model deprecation policy and recommended replacement models, see Model Support Policy.

The following models are deprecated from the Model Catalog:

• Meta Llama 3.1 8B-Instruct
• Mistral NeMo

Migrate to Llama 3.3 70B-Instruct (llama3.3-70b-instruct) and gpt-oss-20b (openai-gpt-oss-20b) models respectively, to avoid service disruption.

The following client libraries for DigitalOcean AI Inference Hub are now available in the official DigitalOcean SDKs. You can use the SDKs to manage serverless and dedicated inference:

• The Python client library is now available in the official DigitalOcean Python client library PyDo. For more information, see the following reference documentation:

  • Serverless Inference

  • Dedicated Inference

• The TypeScript client library is now available in the official DigitalOcean TypeScript library DoTs.

• The official Go client library is available at Gradient Go library.

The Gradient™ SDK will be deprecated in a future release.

The following client libraries for DigitalOcean Gradient™ AI Platform are now available in the official DigitalOcean SDKs. You can use the SDKs to manage DigitalOcean Gradient™ AI Platform resources, including knowledge bases and generative AI agents, and agent, serverless, and dedicated inference:

• The Python client library is now available in the official DigitalOcean Python client library PyDo. For more information, see the following reference documentation:

  • DigitalOcean Gradient™ AI Platform

  • Agent Inference

  • Serverless Inference

  • Dedicated Inference

• The TypeScript client library is now available in the official DigitalOcean TypeScript library DoTs.

• The official Go client library is available at Gradient Go library.

The Gradient™ SDK will be deprecated in a future release.

The following Arcee model is now available on DigitalOcean Gradient™ AI Platform for serverless inference and Agent Development Kit:

• Trinity Large (Public Preview)

For more information, see the Available Models page.

The following Arcee model is now available on DigitalOcean AI Inference Hub for serverless inference:

• Trinity Large (Public Preview)

For more information, see the Available Models page.

NVIDIA B300 GPUs are now available as single-node GPU worker nodes in DigitalOcean Kubernetes (DOKS), by contract only. To add B300 GPU nodes to your cluster, contact sales. Learn more about GPU worker nodes.

Control plane firewalls for DigitalOcean Kubernetes are now in general availability. Control plane firewalls restrict access to your cluster’s API server to a set of allowed IP addresses. Worker node IPs are automatically kept in sync as nodes scale up or down.

You can enable control plane firewalls using the DigitalOcean API, doctl, or Terraform.

NVIDIA B300 GPUs are now generally available in RIC1, by contract only. B300 GPUs are available in 1- and 8-GPU configurations for GPU Droplets via the control panel and via the API using slugs gpu-b300x1-288gb (1 GPU) and gpu-b300x8-2304gb (8 GPUs). Learn more about GPU Droplet plans.

Updated NVIDIA AI/ML Ready (gpu-h100x1-base, gpu-h100x8-base) Droplet base images are now available in the Control Panel and through the API. The image includes DOCA 2.9.3, CUDA 13.1, CUDA drivers 590 (upgraded from Mellanox 23.10-4.0.9.1, Cuda 12.9 and NVIDIA drivers 575), providing the latest NVIDIA GPU support for AI/ML workloads. We’ve also updated the underlying Linux kernel version from 5.15.0-113-generic to 5.15.0-170-generic.

Despite their slug names, these images are compatible with all NVIDIA GPU Droplet types we offer.

Cloud Security Posture Management (CSPM) is now generally available. CSPM evaluates your DigitalOcean resources for misconfigurations and security risks, surfaces findings by severity, and provides guided remediation to help you resolve them. For more information, see the CSPM documentation.

Private Droplets are now in public preview. Private Droplets have no direct public connectivity by default, using VPC-only networking with automatic integration with VPC NAT gateway, VPC peering, and VPC private DNS.

All customers can opt in from the Feature Preview page. Create Private Droplets by setting public_networking: false in the Create Droplet API.

The following OpenAI models are now available on DigitalOcean Gradient™ AI Platform for serverless inference and Agent Development Kit:

• GPT-5.4 mini
• GPT-5.4 nano
• GPT-5.4 pro
• GPT Image 1.5

For more information, see the Available Models page.

The following OpenAI models are now available on Inference for serverless inference:

• GPT-5.4 mini
• GPT-5.4 nano
• GPT-5.4 pro
• GPT Image 1.5

Automatic garbage collection for DigitalOcean Container Registry (DOCR) is now available in public preview. When enabled, DOCR automatically cleans up unreferenced image layers in the background, freeing storage without requiring manual garbage collection runs or read-only downtime. For more details, see Automatic Garbage Collection.

Now in private preview, App Platform’s Scale to Zero feature automatically puts unused web service components to sleep after a configurable period of inactivity and wakes them when they receive external traffic. This helps reduce costs for web services with periods of low or no traffic.

The following NVIDIA model is now available on DigitalOcean Gradient™ AI Platform for serverless inference, Agent Development Kit, and agents:

• Nemotron-3-Super-120B (Public Preview)

For more information, see the Available Models page.

The following NVIDIA model is now available on DigitalOcean AI Inference Hub for serverless inference:

• Nemotron-3-Super-120B (Public Preview)

For more information, see the Available Models page.

We have launched the Richmond, Virginia, USA (ric1) datacenter, which supports GPU Droplets, Kubernetes, and many other products. Learn more in the regional availability matrix.

Gradient AI Dedicated Inference Service is a managed LLM hosting service for optimized inference on dedicated GPUs, now available in public preview and enabled for all users. For more information, see Use Dedicated Inference.

DigitalOcean AI Inference Hub is now available in private preview and is enabled for all users. Inference Hub provides access to a catalog of foundation models with support for serverless inference and dedicated inference, along with a Model Playground for testing models before deployment.

During the private preview period, features and model availability may change.

NVIDIA B300 GPUs are now available in RIC1 as a private preview, by contract only. B300 GPUs are available in 1- and 8-GPU configurations for GPU Droplets via the control panel and via the API using slugs gpu-b300x1-288gb (1 GPU) and gpu-b300x8-2304gb (8 GPUs). Learn more about GPU Droplet plans.

AMD Instinct MI350X GPUs are now available in RIC1 by contract only in 1- and 8-GPU configurations for single- and multi-node GPU Droplets. To create GPU Droplets with MI350X GPUs, contact sales. Learn more about GPU Droplet plans.

Team owners and resource modifiers can now view resource usage and limits in the DigitalOcean Control Panel. You can use this interface to understand resource capacity, manage resource growth, and initiate support requests to increase limits when needed. For more information, see View Resource Limits.

Namespace access keys are now available for Functions. They provide user-specific credentials per namespace, so you can create a key for each user or application and revoke access individually. Keys linked to removed team members are revoked automatically. The legacy shared namespace token is deprecated and will be removed on 3 June 2026. During the migration period, both methods work. After 3 June 2026, legacy tokens will no longer authenticate.

Visit How to Manage Namespace Access Keys to learn more about managing namespace access keys.

Network File Storage now offers two performance tiers. The new Standard Tier provides cost-effective shared storage for general-purpose workloads. The existing High Performance Tier delivers higher throughput that scales with share size for data-intensive workloads like AI/ML training and high-throughput analytics.

Network File Storage is generally available in ATL1, NYC2, and AMS3. For details, see NFS features and pricing.

The following models are now available on DigitalOcean Gradient™ AI Platform for serverless inference, Agent Development Kit, and creating agents:

• GLM 5

• Kimi K2.5

• MiniMax M2.5 (Public Preview)

For more information, see the Available Models page.

Teams and organizations now support dedicated billing and security contact emails. Team and organization owners can add email addresses to receive billing or security notifications independently from the default team or organization contact email. See team contact emails and organization contact emails.

The following OpenAI model is now available on DigitalOcean Gradient™ AI Platform for serverless inference and Agent Development Kit:

• GPT-5.4

For more information, see the Available Models page.

Network File Storage is now available in the AMS3 datacenter.

The routing agent now supports anchor gateway routing, allowing you to use an anchor IP as the source address for outbound traffic from your DOKS cluster.

We now support prompt caching for the following OpenAI models in serverless inference chat completions and responses API:

• GPT-5.3-Codex
• GPT-5.2
• GPT-5.2 pro
• GPT-5.1-Codex-Max
• GPT-5
• GPT-5 mini
• GPT-5 nano
• GPT-4.1
• GPT-4o
• GPT-4o mini
• o1
• o3
• o3-mini
• GPT-image-1

Ubuntu 25.04 has reached end of life. Per our image deprecation policy, this image is available exclusively via the API for the next 30 days before we remove it from our platform.

The DigitalOcean Cloud Security Posture Management (CSPM) dashboard is now in public preview, allowing you to scan your DigitalOcean resources for cloud misconfigurations and security liabilities. The dashboard lists issues by severity and provides mitigation steps on how to resolve them. For more details such as pricing, see our CSPM product page.

The following Anthropic model is deprecated from DigitalOcean Gradient™ AI Platform:

• Claude 3 Opus

Migrate to a supported active model to avoid service disruption. For information on our model deprecation policy and recommended replacement models, see Model Support Policy.

The following OpenAI models are now available on DigitalOcean Gradient™ AI Platform for serverless inference and Agent Development Kit:

• GPT-5.3-Codex

For more information, see the Available Models page.

DigitalOcean Managed PostgreSQL Standard Edition and MySQL Standard Edition clusters now support custom CNAME records, allowing clients to connect through your own hostname instead of the default *.db.ondigitalocean.com address. Custom CNAMEs are available through the API when creating clusters or replicas. Advanced Edition clusters do not support custom CNAMEs. For more information, see Configure Custom CNAMEs for PostgreSQL or Configure Custom CNAMEs for MySQL.

You can now use reasoning with serverless inference chat completion. For more information, see Use Reasoning.

An updated AMD AI/ML Ready (gpu-amd-base) Droplet base image is now available in the Control Panel and through the API. The image includes ROCm 7.0.2 (upgraded from 6.4.3), providing the latest AMD GPU support for AI/ML workloads.

New MI300X Droplets may encounter firmware incompatibilities with this image. If you experience issues, contact support. We can provide a temporary snapshot to help you proceed. For new MI300X deployments, consider snapshotting an existing AMD GPU Droplet created before 13 February 2026 as a fallback.

A Fedora 43 (fedora-43-x64) Droplet base image is now available in the Control Panel and through the API.

The following Anthropic models are deprecated from DigitalOcean Gradient™ AI Platform:

• Claude 3.5 Haiku
• Claude 3.7 Sonnet
• Claude 3.5 Sonnet

Migrate to a supported active model to avoid service disruption. For information on our model deprecation policy and recommended replacement models, see Model Support Policy.

OpenAI and Anthropic commercial models now default to using DigitalOcean API keys when creating new agents. This allows you to have consolidated billing for all agent usage and no keys to manage on your own. If you want, you can bring your own keys when creating new agents or continue using your own keys for existing agents. For more information on pricing, see the pricing details.

AMD Instinct MI350X GPUs are now available in ATL1 by contract only in 1- and 8-GPU configurations for single- and multi-node GPU Droplets. To create GPU Droplets with MI350X GPUs, contact sales. Learn more about GPU Droplet plans.

The following Anthropic model is now available on DigitalOcean Gradient™ AI Platform for serverless inference, Agent Development Kit, and creating agents:

• Claude Sonnet 4.6

For more information, see the Available Models page.

DigitalOcean Functions now supports functions written in Python 3.12 and Python 3.13.

As of 1 February 2026, DigitalOcean charges Retail Sales Tax (RST) for customers with a tax location in Manitoba. This complies with Information Bulletin – RST 033 issued by the Manitoba Government.

Services provided by DigitalOcean to customers located in Manitoba are subject to 7% RST in addition to 5% GST for a total tax rate of 12%. Learn more about taxes in Canada.

DigitalOcean Functions now supports functions written in PHP 8.3, PHP 8.4, and PHP 8.5.

DigitalOcean Functions now supports functions written in Go 1.24 and Go 1.25.

Multimodal models for image and audio generation, provided by fal, are now in general availability. You can use these models for serverless inference. For examples of how to use these models, see Generate Image, Audio, or Text-to-Speech Using fal Models.

The App Platform UI now includes a Networking tab where you can manage domains, redirects, request routes, dedicated egress IPs, and VPC network settings in one place.

App Platform’s Rust buildpack is now available, simplifying the build and deployment of Rust applications on App Platform.

We have updated the following buildpacks:

• Node.js buildpack: We have added new Node.js runtimes to address CVEs included in the January 2026 security releases.
  • Node.js 25.3.0, 25.4.0
  • Node.js 24.12.0, 24.13.0
  • Node.js 22.22.0
  • Node.js 20.20.0

End users and agent developers can now provide feedback on the quality and helpfulness of agent responses. The feedback is collected through the chatbot interface, agent playground, and log stream traces and stored in the traces. For more information, see Provide Agent Feedback and View Conversation Logs, Traces, and Insights.

The following Anthropic and OpenAI models are now available on DigitalOcean Gradient™ AI Platform:

• Haiku 4.5 for serverless inference, ADK, and agent creation

• GPT-5.2 pro for serverless inference, ADK, and agent creation

• Opus 4.6 for serverless inference and ADK

For more information, see the Available Models page.

We have enabled trace storage by default for both newly created and existing agents.

DigitalOcean Functions now supports functions written in Node.js 22 and Node.js 24.

We now support prompt caching for the following Anthropic models:

• Claude Sonnet 4.5
• Claude Sonnet 4
• Claude Opus 4.5
• Claude Opus 4.1
• Claude Opus 4

Using prompt caching with serverless inference chat completion significantly reduces the cost for inference.

The following OpenAI models are now available on DigitalOcean Gradient™ AI Platform for serverless inference, Agent Development Kit, and creating agents:

• GPT-5.2

• GPT-5.1-Codex-Max

For more information, see the Available Models page.

The daft package in the Functions Python 3.11 runtime has been updated from version 0.4.8 to 0.4.10. Visit the Python 3.11 runtime documentation for more information about included packages.

We have enabled silent re-authentication for single sign-on (SSO) when using the DigitalOcean Control Panel. DigitalOcean automatically refreshes the identity provider (IdP) access tokens, so you are no longer unexpectedly logged out when the tokens expire. This improves security and minimizes interruptions to your workflow.

App Platform’s .NET buildpack is now available, enabling simplified build and deployment of .NET applications on App Platform.

AMD Instinct MI325X GPUs are now available in SFO3 by contract only in 1- and 8-GPU configurations for single- and multi-node GPU Droplets. To create GPU Droplets with MI325X GPUs, contact sales. Learn more about GPU Droplet plans.

To comply with the City of Chicago’s Personal Property Lease Transaction Tax, the sales tax rate for customers in Chicago has increased from 11% to 15%. Learn more about taxes in the United States of America.

Paperspace support emails now use the digitalocean.com domain. Add digitalocean.com to your email allowlist and reply only to emails from digitalocean.com. Use the existing Paperspace Support form to open new Paperspace support tickets.

We have updated the following buildpacks for App Platform:

• Go buildpack: We have added the following Go versions. If you have an existing Go app that uses v0, we recommend upgrading to v1.
  • go1.23.10 - go1.23.12
  • go1.24.4 - go1.24.11
  • go1.25.0 - go1.25.5
• PHP buildpack: We have added the following PHP versions. If you have an existing PHP app that uses v1, we recommend upgrading to v2.
  • PHP 8.1.33
  • PHP 8.2.29
  • PHP 8.3.23 - 8.3.28
  • PHP 8.4.10 - 8.4.15
• Python buildpack: We have added the following Python versions. If you are on Ubuntu-22 and have an existing Python app that is on v3, v2, v1 or v0, we recommend upgrading to v4.
  • Python 3.14.0 - 3.14.1
  • Updated uv to 0.9.14
• Node.JS buildpack: We have added the following Nodejs versions. Visit the Node.js buildpack to learn more about specifying a Node.js Engine version.
  • Node.js 20.19.3 - 20.19.6
  • Node.js 22.17.0 - 22.21.1
  • Node.js 24.3.0 - 24.11.1
  • Node.js 25.0.0 - 25.2.1
• Ruby buildpack: We have updated the default Ruby version to 3.3.9 and have added the following Ruby versions. If you have an existing Ruby app that uses v1, we recommend upgrading to v2.
  • Ruby 3.2.9
  • Ruby 3.3.9 - 3.3.10
  • Ruby 3.4.5 - 3.4.7
  • Ruby 4.0.0-preview2

To comply with the Republic of Kazakhstan’s new tax code, the VAT rate for customers in Kazakhstan has increased from 12% to 16%. Learn more about taxes in Kazakhstan.

To comply with the Government of the Russian Federation’s 2026–28 budget, the VAT rate for customers in Russia has increased from 20% to 22%. Learn more about taxes in Russia.

Starting 1 January 2026, CPU Droplets are billed per-second, with a minimum charge of 60 seconds or $0.01, whichever is higher.