How to use Continue with a self-hosted model on a cloud GPU
Rent a GPU on Deploy, then open a terminal in JupyterLab or over SSH and serve an open-weight model there. Install Continue the way the official install page prints, put the documented config.yaml on the machine that runs the editor, and set apiBase to the forwarded address of that server. Copy anything the agent wrote on the instance before you destroy it. Destroying the instance deletes its disk and stops the server.
An agent can rent the GPU and serve the model the editor calls. The agent pages are GPU compute for AI agents and the agent integration guide. Those pages are how an agent discovers a GPU and runs work on it. This page is the Continue setting. An editor client whose install page is the VS Code extension is How to use Cline with a self-hosted model on a cloud GPU. A terminal client whose OpenAI-compatible page sets OPENAI_API_BASE is How to use Aider with a self-hosted model on a cloud GPU. A local agent UI whose Docker command publishes port 3000 is How to run OpenHands with a self-hosted model on a cloud GPU. A coding-agent server whose card prints --tool-call-parser qwen3_coder is How to serve Qwen3-Coder for coding agents on a cloud GPU. A general vLLM install is How to serve an OpenAI-compatible API with vLLM on a cloud GPU. Ollama’s install and port 11434 are How to run Ollama on a cloud GPU. An embeddings server is How to serve embeddings and rerankers with Text Embeddings Inference on a cloud GPU. The Cursor fields on How do I use CloudGPU as a custom base URL in Cursor? are for the hosted gateway, not for the localhost URL of a server you start. How published weight sizes are read: How much VRAM do I need to run an LLM?. The tunnel is How to use SSH port forwarding on a cloud GPU. If the server runs out of memory: How to fix CUDA out of memory on a cloud GPU. If the server raises no kernel image on an RTX 5090: How to fix CUDA no kernel image on an RTX 5090.
The Continue LICENSE file begins with “Apache License Version 2.0, January 2004”. The appendix copyright line is “Copyright 2023 Continue Dev, Inc.” The README license line is “Apache 2.0 © 2023-2026 Continue Dev, Inc.” The GitHub license field on 2026-11-14 was Apache-2.0. That license text is not a CloudGPU price.
The steps below are copied from the upstream pages named in each step. They were not run on a CloudGPU instance for this page. Anything this page adds around them is marked. The provider pages also name other providers. This page does not list those names.
What the docs print
The README at commit 5522c6f44ca0ac3528b37244818fbfa39b5af470 (committed 2026-07-21T04:00:09Z) says the continuedev/continue repository is no longer actively maintained and is read-only for all users. It says Continue is available as a CLI, a VS Code extension, and a JetBrains plugin, and that the final release is 2.0.0. A note under the JetBrains heading says “We recommend using the Continue CLI instead of the JetBrains plugin.” The README does not print an npm install command. It links the npm package @continuedev/cli. The repository pushed_at on 2026-11-14 was 2026-11-13T13:26:06Z. This page follows the install page and the config pages at that commit. It did not install the CLI.
The install page at that commit documents two IDE tabs. The VS Code tab says to click Install on the Visual Studio Marketplace page for Continue.continue, then click Install again inside VS Code. The JetBrains tab says to open Settings with Ctrl + Alt + S, select Plugins, search for “Continue”, and click Install. The live install URL returned HTTP 200 on 2026-11-14 and still contained those strings.
The YAML migration page says to create config.yaml in the Continue Global Directory, “~/.continue on Mac, %USERPROFILE%\.continue”. It says that if config.yaml is present, it is loaded instead of config.json. That sentence does not print a Linux path. This page does not add one. The editor config lives on the computer that runs Continue. The model server is the process on the instance.
The roles introduction lists chat, autocomplete, edit, apply, embed, and rerank. It says roles are set on a config.yaml model block, and that config.json does not define roles explicitly. The config.yaml reference says roles can include chat, autocomplete, embed, rerank, edit, apply, and summarize, and that the default is [chat, edit, apply, summarize]. The same bullet says the summarize role is not currently used. Required model properties on that page are name, provider, and model. apiBase “can be used to override the default API base”.
Choosing a GPU
The install page, the config reference, and the OpenAI-compatible, Ollama, and vLLM provider pages print no GPU GB figure. The Ollama guide prints RAM lines: “RAM: Minimum 8GB (16GB+ recommended)”, and a later list of “1.5B-3B models: ~4GB RAM”, “7B models: ~8GB RAM”, “13B models: ~16GB RAM”, and “32B models: ~32GB RAM”. Those lines say RAM. This page does not treat them as vram_gb and does not mark a catalog row as enough. The same guide also prints a storage line. This page does not copy it.
GET /v1/public/compute-capabilities does not list GPUs. It returns feature flags. The memory column is vram_gb from gpu-prices.json, updated_at 2026-11-14T02:04:03.491Z. GET /v1/public/gpu-catalog the same day (generated_at 2026-11-14T02:04:03.382Z) listed the same names and the same vram_gb values, plus Ascend 910B at 64, which is not in gpu-prices.json and is not listed here. The catalog response also includes an availability field. This page does not copy it. Every prices row has gpu_count 1. Every catalog row has min_gpu_count 1. Which listing you can start is on Deploy.
The catalog template list had no Continue template. It did have Ollama templates. Those templates are a different install. This page does not treat them as Continue.
| Catalog name | vram_gb |
|---|---|
| RTX 2080 | 8 |
| RTX 3080 | 10 |
| RTX 2080 Ti | 11 |
| RTX 3080 Ti | 12 |
| V100 16G | 16 |
| A30 | 24 |
| RTX 3090 | 24 |
| RTX 4090 | 24 |
| RTX 4090D | 24 |
| RTX 4080 SUPER 32G | 32 |
| RTX 5090 | 32 |
| RTX 5090D | 32 |
| V100 32G | 32 |
| A100 40G | 40 |
| A800 40G | 40 |
| A40 | 48 |
| RTX 4090 48G | 48 |
| RTX 4090D 48G | 48 |
| RTX A6000 | 48 |
| NVIDIA RTX PRO 5000 72G | 72 |
| RTX PRO 5000 72G | 72 |
| A100 80G | 80 |
| A800 80G | 80 |
| H800 | 80 |
| RTX 6000D | 83 |
| H20 96G | 96 |
| RTX PRO 6000 96G | 96 |
| RTX PRO 6000 Max-Q | 96 |
| H20-3e 141G | 141 |
Install Continue
The install page’s VS Code steps, copied. Click Install on the Marketplace page, then click Install again in VS Code. The page says the Continue logo appears on the left sidebar.
The JetBrains steps, copied. Open Settings with Ctrl + Alt + S, search Plugins for “Continue”, and click Install. The README at this commit still says to prefer the CLI over that plugin. This page did not install either one. Continue itself is the editor extension. The weights are whatever model server you start on the instance. This page does not clone the Continue repository onto the GPU.
Point chat, autocomplete, and embeddings at the server
The OpenAI-compatible section of docs/customize/model-providers/top-level/openai.mdx says that for an OpenAI-compatible provider you change apiBase. Its YAML tab, limited to the fields this page keeps:
name: My Config
version: 0.0.1
schema: v1
models:
- name: <OPENAI_API_COMPATIBLE_PROVIDER_MODEL>
provider: openai
model: <MODEL_NAME>
apiBase: http://localhost:8000/v1
apiKey: <YOUR_CUSTOM_API_KEY>
The same file’s first YAML example does not set apiBase. This page follows the compatible example. The page also names other providers. This page does not list them. A JSON tab on the same page is labeled “JSON (Deprecated)”. The migration page says YAML is loaded instead of config.json when both exist.
The self-host guide says that when the API is OpenAI-compatible, use the “openai” provider and change the “baseUrl”. The YAML blocks on the provider pages use the key apiBase. This page does not treat baseUrl and apiBase as the same key. The self-host guide’s Ollama example sets model: llama2-7b and apiKey. It does not set apiBase.
The Ollama provider page’s YAML, copied. The comment is the page’s comment:
name: My Config
version: 0.0.1
schema: v1
models:
- name: <MODEL_NAME>
provider: ollama
model: <MODEL_ID>
apiBase: http://<my endpoint>:11434 # if running a remote instance of Ollama
The vLLM provider page says to run vllm serve and points at the vLLM server docs. Its chat, autocomplete, and embeddings YAML tabs set provider: vllm and an apiBase that ends in /v1. The chat example model id is meta-llama/Meta-Llama-3.1-8B-Instruct and the apiBase placeholder is http://<vllm chat endpoint>/v1. The autocomplete example model id is Qwen/Qwen2.5-Coder-1.5B, with roles: [autocomplete]. The embeddings example model id is nomic-ai/nomic-embed-text-v1, with roles: [embed]. The embeddings block’s name on that page is “VLLM Nomad Embed Text”. This page did not download those repositories.
For a vLLM or Ollama process on the instance, apiBase is the client address that server’s guide prints, after the port is forwarded. How to use SSH port forwarding on a cloud GPU is how a program on your computer reaches a port on the instance. This page does not add an ssh command. The hosted gateway at https://cloudgpu.app/v1 is the Cursor guide, not this server.
The Ollama guide’s remote line says you can update apiBase with the IP address of a remote machine serving Ollama, and one YAML on that guide sets apiBase: http://localhost:11434. A later troubleshooting line says to set OLLAMA_HOST=0.0.0.0:11434 for remote connections. Those are the guide’s lines. This page did not set them.
The roles introduction says you pick the active model for each role in the editor: above the main input, click the 3 dots, then the cube icon, and use the dropdowns. This page did not click those controls.
When it fails
CUDA out of memory. The Continue pages this page read print RAM lines and no GPU GB figure. If the model server raises CUDA out of memory, use how to fix CUDA out of memory on a cloud GPU. This page does not pick a smaller model id for you.
“Model requires more system memory”. The Ollama provider page says Continue may set a higher default context length than other Ollama tools. The documented example sets defaultCompletionOptions.contextLength to 2048 on model: deepseek-r1:latest. That 2048 is the example on that page. This page did not save it.
“CUDA error: no kernel image is available for execution on the device” on an RTX 5090. The Continue pages this page read do not pin a PyTorch build and do not name compute capability 12.0. The server you start might. The RTX 5090 guide says that card is compute capability 12.0 (sm_120) and that PyTorch 2.7 and later, on a build for CUDA 12.8 or newer, is the line that announced support. This page did not install a torch wheel and did not re-test a server on an RTX 5090. Catalog RTX 5090 and RTX 5090D are 32. That is the catalog number only.
What to copy off, then destroy
The Continue pages do not name an output directory on the GPU. Chat transcripts stay in the editor. config.yaml stays on the computer that runs the editor. Any file the agent writes on the instance is on that disk. Copy it before you destroy the instance. Save your work before you destroy a cloud GPU is the checklist. The terms say data on local instance storage is permanently deleted when the instance is destroyed. The server stops with the instance.
The pricing page FAQ states the billing rule in these words:
Per-minute options have a 1-minute minimum; unused prepaid minutes are refunded when you destroy the instance. Some configurations use whole-hour billing with a 1-hour minimum; unused time within a charged hour is not refunded on voluntary stop. The billing period, minimum charge and initial prepayment are shown before confirmation. No subscription.
Prices are not copied here.
Frequently asked questions
What memory figure do the Continue docs print?
The install page, the config.yaml reference, and the OpenAI-compatible, Ollama, and vLLM provider pages print no GPU GB figure. The Ollama guide prints RAM: Minimum 8GB (16GB+ recommended), then ~4GB RAM for 1.5B-3B models, ~8GB RAM for 7B models, ~16GB RAM for 13B models, and ~32GB RAM for 32B models. Those lines say RAM. This page does not mark a catalog row as enough. Which listing you can start is on https://cloudgpu.app/deploy.
Which install steps and which config file do the docs print?
The install page VS Code steps are click Install on the Marketplace page for Continue.continue, then click Install again in VS Code. The JetBrains steps are Settings, Plugins, search Continue, Install. The README says the repository is read-only, the final release is 2.0.0, and it recommends the CLI instead of the JetBrains plugin. The migration page says to create config.yaml in the Continue Global Directory, ~/.continue on Mac or %USERPROFILE%\.continue, and that config.yaml is loaded instead of config.json when it is present. That page does not print a Linux path.
How do chat, autocomplete, and embeddings reach a server on the rented GPU?
The roles introduction lists chat, autocomplete, edit, apply, embed, and rerank. The OpenAI-compatible YAML sets provider to openai, an apiBase of http://localhost:8000/v1, and an apiKey. The Ollama YAML sets provider to ollama and an apiBase on port 11434. The vLLM page sets provider to vllm, an apiBase that ends in /v1, roles autocomplete for the Qwen2.5-Coder 1.5B example, and roles embed for the nomic-embed-text example. The config file is on the computer that runs the editor. The SSH port-forwarding guide is how that computer reaches the instance port. This page did not open a tunnel.
What license does the file print?
The Continue LICENSE file begins with Apache License Version 2.0, January 2004. The appendix copyright line is Copyright 2023 Continue Dev, Inc. The README license line is Apache 2.0 © 2023-2026 Continue Dev, Inc. The GitHub license field on 2026-11-14 was Apache-2.0. That license text is not a CloudGPU price.
Sources (read 2026-11-14, Asia/Shanghai)
- continuedev/continue README (read-only repository, CLI / VS Code / JetBrains, final 2.0.0, CLI preferred over the JetBrains plugin, Apache 2.0 line): commit
5522c6f44ca0ac3528b37244818fbfa39b5af470, committed 2026-07-21T04:00:09Z. GitHub license fieldApache-2.0.pushed_at2026-11-13T13:26:06Z docs/ide-extensions/install.mdx(VS Code MarketplaceContinue.continue, JetBrains SettingsCtrl+Alt+S): same commit. Live URL https://docs.continue.dev/ide-extensions/install returned HTTP 200 on 2026-11-14docs/reference.mdx(config.yaml, requiredname/provider/model,apiBase, roles includingsummarizemarked not currently used, default[chat, edit, apply, summarize]): same commitdocs/reference/yaml-migration.mdx(Global Directory~/.continueon Mac,%USERPROFILE%\.continue,config.yamlloaded instead ofconfig.json): same commitdocs/customize/model-roles/intro.mdx(chat,autocomplete,edit,apply,embed,rerank): same commitdocs/customize/model-providers/top-level/openai.mdx(OpenAI-compatibleapiBase: http://localhost:8000/v1,provider: openai): same commit. That page also names other providers. This page does not list themdocs/customize/model-providers/top-level/ollama.mdx(provider: ollama, port 11434, “Model requires more system memory”, examplecontextLength: 2048): same commitdocs/customize/model-providers/more/vllm.mdx(provider: vllm, chat / autocomplete / embed examples,roles): same commitdocs/guides/how-to-self-host-a-model.mdx(prose key “baseUrl”, examplemodel: llama2-7b): same commitdocs/guides/ollama-guide.mdx(RAM lines,OLLAMA_HOST=0.0.0.0:11434): same commit. This page does not copy that guide’s storage line- Public GPU memory: https://cloudgpu.app/gpu-prices.json (
updated_at2026-11-14T02:04:03.491Z) and https://cloudgpu.app/v1/public/gpu-catalog (generated_at2026-11-14T02:04:03.382Z).GET /v1/public/compute-capabilitiesreturned feature flags and no GPU list. The catalog had no Continue template
Cards you can start are on the Deploy page. Prices are on https://cloudgpu.app/pricing and are not repeated here.
Last updated: 2026-11-14 (Asia/Shanghai).