Skip to content

Ollama Cloud Models #6582

Closed
Closed
Ollama Cloud Models#6582
Assignees
rekram1-node
Labels
bugSomething isn't working

Description

@mdlmarkham
mdlmarkham
opened on Jan 1, 2026

Description

I'm using Ollama Cloud with OpenCode - it works great, except that when new models are added to the service they don't show up in OpenCode. Is there a way to force a refresh of the available models?

OpenCode version

1.0.223

Steps to reproduce

Connect OpenCode to OllamaCloud as a provider using an API key.
Code...be happy....
Observe that OllamaCloud has released a new model.
Try to add it using /models
Observe that it isn't available in OpenCode
Be sad.

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

Activity

  1. mdlmarkham
    added
    bugSomething isn't working
    on Jan 1, 2026
  2. github-actions
    assigned
    rekram1-node
    on Jan 1, 2026

  3. github-actions commented on Jan 1, 2026

    @github-actions
    github-actionsBot
    on Jan 1, 2026 – with GitHub Actions
    Contributor

    This issue might be a duplicate of existing issues. Please check:

    These issues address the core problem of models not showing up in OpenCode and suggest solutions including manual refresh and auto-discovery features. Feel free to ignore if your specific case is different.

  4. mdlmarkham commented on Jan 1, 2026

    @mdlmarkham
    mdlmarkham
    on Jan 1, 2026
    Author

    I'm trying to work around this by configuring an alternative provider in opencode.json, but so far I can't get it to work:

    "provider": {
"ollamacloudManual": {
  "npm": "@ai-sdk/openai-compatible",
  "name": "OllamaCloudManual",
  "options": {
    "baseURL": "https://ollama.com",
    "apiKey": "asdfasdfasdfasdfasdfasdfasdfasdfasdf"
  },
  "models": {
    "devstral-2:123b-cloud": {
      "name": "devstral 2 123B Cloud"
    },
    "devstral-2:latest": {
      "name": "devstral 2 123B Cloud - Latest"
    },
    "minimax-m2.1:cloud": {
      "name": "MiniMax M2.1 Cloud"
    }
  }
}

    }

  5. rekram1-node commented on Jan 1, 2026

    @rekram1-node
    rekram1-node
    on Jan 1, 2026
    Contributor

    u dont need a new provider u can override or add models for the existing one, the models list is here:
    https://models.dev/

    We should use the provider /models endpoints to populate the list automatically too, but normally they dont provide enough information to populate model list correctly besides the name and id of the model

  6. mdlmarkham commented on Jan 1, 2026

    @mdlmarkham
    mdlmarkham
    on Jan 1, 2026
    Author

    Thanks @rekram1-node. In this case, the models that I would like to use are being made available by OllamaCloud (https://ollama.com/library/devstral-2) but aren't visible on models.dev

    What's the syntax for the override?

  7. spiffytech commented on Jan 4, 2026

    @spiffytech
    spiffytech
    on Jan 4, 2026

    It seems like opencode pulls the list of available models from models.dev. So as long as that's missing Minimax M2.1, GLM 4.7, etc., so will opencode.

    I added these models locally, like so:

    {
  "$schema": "https://opencode.ai/config.json",
  "permission": {
    "edit": "ask",
    "bash": "ask",
    "skill": "ask",
    "webfetch": "ask",
    "doom_loop": "ask",
    "external_directory": "ask"
  },
  "provider": {
    "ollama-cloud": {
      "name": "Ollama Cloud",
      "models": {
        "minimax-m2.1": {
          "name": "Minimax M2.1",
          "limit": { "context": 204800, "output": 0 }
        },
        "glm-4.7": {
          "name": "GLM 4.7",
          "limit": { "context": 202752, "output": 0 }
        }
      }
    }
  }
}

    I got the context lengths by talking to the Ollama Cloud API:

    $ curl https://ollama.com/api/show -d '{
    "model": "minimax-m2.1"
  }' | jq '.'

{
  "details": {
    "parent_model": "minimax-m2.1",
    "format": "",
    "family": "minimax-m2",
    "families": null,
    "parameter_size": "230000000000",
    "quantization_level": "FP8"
  },
  "model_info": {
    "general.architecture": "minimax-m2",
    "general.parameter_count": 230000000000,
    "minimax-m2.context_length": 204800,
    "minimax-m2.embedding_length": 3072
  },
  "capabilities": [
    "completion",
    "tools",
    "thinking"
  ],
  "modified_at": "0001-01-01T00:00:00Z"
}

    And the list of available models:

    $ curl -s https://ollama.com/api/tags | jq -r '.models[] | .name' | sort

cogito-2.1:671b
deepseek-v3.1:671b
deepseek-v3.2
devstral-2:123b
devstral-small-2:24b
gemini-3-flash-preview
gemini-3-pro-preview
gemma3:12b
gemma3:27b
gemma3:4b
glm-4.6
glm-4.7
gpt-oss:120b
gpt-oss:20b
kimi-k2:1t
kimi-k2-thinking
minimax-m2
minimax-m2.1
ministral-3:14b
ministral-3:3b
ministral-3:8b
mistral-large-3:675b
nemotron-3-nano:30b
qwen3-coder:480b
qwen3-next:80b
qwen3-vl:235b
qwen3-vl:235b-instruct
rnj-1:8b

  8. mdlmarkham commented on Jan 7, 2026

    @mdlmarkham
    mdlmarkham
    on Jan 7, 2026
    Author

    Thanks @spiffytech - I'll give it a try!

  9. better319 commented on Jan 12, 2026

    @better319
    better319
    on Jan 12, 2026

    u dont need a new provider u can override or add models for the existing one, the models list is here: https://models.dev/

    We should use the provider /models endpoints to populate the list automatically too, but normally they dont provide enough information to populate model list correctly besides the name and id of the model

    https://models.dev/ not the latest models，justlike ollama cloud glm4.7 m2.1 isready

  10. better319 commented on Jan 12, 2026

    @better319
    better319
    on Jan 12, 2026

    It seems like opencode pulls the list of available models from models.dev. So as long as that's missing Minimax M2.1, GLM 4.7, etc., so will opencode.

    I added these models locally, like so:

    {
    "$schema": "https://opencode.ai/config.json",
    "permission": {
    "edit": "ask",
    "bash": "ask",
    "skill": "ask",
    "webfetch": "ask",
    "doom_loop": "ask",
    "external_directory": "ask"
    },
    "provider": {
    "ollama-cloud": {
    "name": "Ollama Cloud",
    "models": {
    "minimax-m2.1": {
    "name": "Minimax M2.1",
    "limit": { "context": 204800, "output": 0 }
    },
    "glm-4.7": {
    "name": "GLM 4.7",
    "limit": { "context": 202752, "output": 0 }
    }
    }
    }
    }
    }
    I got the context lengths by talking to the Ollama Cloud API:

    $ curl https://ollama.com/api/show -d '{
    "model": "minimax-m2.1"
    }' | jq '.'

    {
    "details": {
    "parent_model": "minimax-m2.1",
    "format": "",
    "family": "minimax-m2",
    "families": null,
    "parameter_size": "230000000000",
    "quantization_level": "FP8"
    },
    "model_info": {
    "general.architecture": "minimax-m2",
    "general.parameter_count": 230000000000,
    "minimax-m2.context_length": 204800,
    "minimax-m2.embedding_length": 3072
    },
    "capabilities": [
    "completion",
    "tools",
    "thinking"
    ],
    "modified_at": "0001-01-01T00:00:00Z"
    }
    And the list of available models:

    $ curl -s https://ollama.com/api/tags | jq -r '.models[] | .name' | sort

    cogito-2.1:671b
    deepseek-v3.1:671b
    deepseek-v3.2
    devstral-2:123b
    devstral-small-2:24b
    gemini-3-flash-preview
    gemini-3-pro-preview
    gemma3:12b
    gemma3:27b
    gemma3:4b
    glm-4.6
    glm-4.7
    gpt-oss:120b
    gpt-oss:20b
    kimi-k2:1t
    kimi-k2-thinking
    minimax-m2
    minimax-m2.1
    ministral-3:14b
    ministral-3:3b
    ministral-3:8b
    mistral-large-3:675b
    nemotron-3-nano:30b
    qwen3-coder:480b
    qwen3-next:80b
    qwen3-vl:235b
    qwen3-vl:235b-instruct
    rnj-1:8b

    THK,URL C:\Users\Administrator.config\opencode\opencode.json
    and terminal:opencode models --refresh
    it's ok now!
    Image

  11. ting-ff commented on Jan 15, 2026

    @ting-ff
    ting-ff
    on Jan 15, 2026

    u dont need a new provider u can override or add models for the existing one, the models list is here: https://models.dev/

    We should use the provider /models endpoints to populate the list automatically too, but normally they dont provide enough information to populate model list correctly besides the name and id of the model

    This would be relatively trivial to implement, right? Just pull both lists and do a disambiguation. It's better to have a complete list of models, with some information missing, than to just omit models without complete information.

  12. github-actions commented on Mar 26, 2026

    @github-actions
    github-actionsBot
    on Mar 26, 2026 – with GitHub Actions
    Contributor

    To stay organized issues are automatically closed after 90 days of no activity. If the issue is still relevant please open a new one.

  13. github-actions
    closed this as completedon Mar 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

bugSomething isn't working

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions