Error in LMStudio models parsing

[jbpin]

Describe the bug / provide steps to reproduce it

Summary:
LMStudio model response cannot be parsed when a model in the list missed the quantization parameter.

Caused by:
    missing field `quantization` at line 24 column 5
2025-01-28T10:13:16.728349+01:00 [ERROR] Unable to parse LM Studio models response

Steps to trigger the problem:

1. Download SmolVLM model in LMStudio
2. Refresh Zed LMStudio with the connect button
3. Open zed log

Actual Behavior:
Zero model are listed from LMStudio because only one is not suitable.
Expected Behavior:
Filter for not compatible models and display the one that match the necessary parameter.

LMStudio response to the models query :

2025-01-28 10:14:38 [DEBUG] 
Received request: GET to /api/v0/models
2025-01-28 10:14:38  [INFO] 
Returning {
  "data": [
    {
      "id": "deepseek-r1-distill-llama-8b",
      "object": "model",
      "type": "llm",
      "publisher": "lmstudio-community",
      "arch": "llama",
      "compatibility_type": "gguf",
      "quantization": "Q8_0",
      "state": "loaded",
      "max_context_length": 131072,
      "loaded_context_length": 4096
    },
    {
      "id": "smolvlm-500m-instruct",
      "object": "model",
      "type": "vlm",
      "publisher": "mlx-community",
      "arch": "idefics3",
      "compatibility_type": "mlx",
      "state": "not-loaded",
      "max_context_length": 8192
    },
    {
      "id": "text-embedding-nomic-embed-text-v1.5",
      "object": "model",
      "type": "embeddings",
      "publisher": "nomic-ai",
      "arch": "nomic-bert",
      "compatibility_type": "gguf",
      "quantization": "Q4_K_M",
      "state": "not-loaded",
      "max_context_length": 2048
    }
  ],
  "object": "list"
}

Zed Version and System Specs

Zed: v0.170.4 (Zed)
OS: macOS 15.2.0
Memory: 128 GiB
Architecture: aarch64

[notpeter]

I was not able to reproduce. Can you provide more details steps?

Here's what I did:

1. LM Studio is running with API access enabled (tested working in Zed from previous setup)
2. Fetch and load these models:

lms get deepseek-r1-distill-llama-8b
lms load deepseek-r1-distill-llama-8b
lms get gaianet/Nomic-embed-text-v1.5-Embedding-GGUF
lms load gaianet/Nomic-embed-text-v1.5-Embedding-GGUF/nomic-embed-text-v1.5.f16.gguf
lms get mlx-community/SmolVLM-500M-Instruct-4bit
lms load mlx-community/SmolVLM-500M-Instruct-4bit

3. Quit and reload Zed.
4. Open Assistant Panel and open model selector.
5. Open Zed log

No errors and models seem to appear.

Perhaps you are loading slightly different models / repacks which have different metadata?
Note I only have 64GB of ram, so LM Studio may be steering me towards models I can run.

[teslanaut]

I've getting this issue since 2 days ago on Mac OS also. Latest Zed + latest LMStudio. Clicking connect in Zed returns the error mentioned above in the Zed logs.

Error persists Zed 0.171.3.

Zed Version and System Specs
Zed: v0.170.4 (Zed)
OS: macOS 15.2.0
Memory: 128 GiB
Architecture: aarch64

[notpeter]

Would love to reproduce this.

@jbpin How did you get smolvlm-500m-instruct without a quantization parameter? (how did you download this).

@teslanaut Do you have a model listed in curl http://localhost:1234/v1/models which doesn't have a quantization field?

[teslanaut]

Here is the output of curl http://localhost:1234/v1/models:

v1/models json

{
  "data": [
    {
      "id": "deepseek-r1-distill-llama-70b",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "text-embedding-nomic-embed-text-v1.5",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "fuseo1-deepseekr1-qwen2.5-coder-32b-preview",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "qwen2.5-14b-deepseek-r1-1m-mlx",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "qwen2.5-14b-instruct-1m",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "zelensky-78b",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "deepseek-r1-distill-qwen-32b",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "pixtral-12b",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "mistral-nemo-instruct-2407",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "qwen2.5-coder-32b-instruct-128k",
      "object": "model",
      "owned_by": "organization_owner"
    },
    {
      "id": "qwen2.5-coder-32b-instruct",
      "object": "model",
      "owned_by": "organization_owner"
    }
  ],
  "object": "list"
}

And my list of models in LM Studio

models.txt

[notpeter]

Ooops. I meant curl http://localhost:1234/api/v0/models

[teslanaut]

Here you go:

model list json

{ "data": [ { "id": "mistral-small-24b-instruct-2501", "object": "model", "type": "llm", "publisher": "unsloth", "arch": "llama", "compatibility_type": "gguf", "quantization": "Q8_0", "state": "not-loaded", "max_context_length": 32768 }, { "id": "text-embedding-nomic-embed-text-v1.5", "object": "model", "type": "embeddings", "publisher": "nomic-ai", "arch": "nomic-bert", "compatibility_type": "gguf", "quantization": "Q4_K_M", "state": "not-loaded", "max_context_length": 2048 }, { "id": "fuseo1-deepseekr1-qwen2.5-coder-32b-preview", "object": "model", "type": "llm", "publisher": "mlx-community", "arch": "qwen2", "compatibility_type": "mlx", "quantization": "4bit", "state": "not-loaded", "max_context_length": 131072 }, { "id": "qwen2.5-14b-deepseek-r1-1m-mlx", "object": "model", "type": "llm", "publisher": "m-i", "arch": "qwen2", "compatibility_type": "mlx", "quantization": "8bit", "state": "not-loaded", "max_context_length": 1010000 }, { "id": "qwen2.5-14b-instruct-1m", "object": "model", "type": "llm", "publisher": "bartowski", "arch": "qwen2", "compatibility_type": "gguf", "quantization": "Q8_0", "state": "not-loaded", "max_context_length": 1010000 }, { "id": "zelensky-78b", "object": "model", "type": "llm", "publisher": "nisten", "arch": "qwen2", "compatibility_type": "gguf", "state": "not-loaded", "max_context_length": 98304 }, { "id": "deepseek-r1-distill-llama-70b", "object": "model", "type": "llm", "publisher": "unsloth", "arch": "llama", "compatibility_type": "gguf", "quantization": "Q4_K_M", "state": "not-loaded", "max_context_length": 131072 }, { "id": "deepseek-r1-distill-qwen-32b", "object": "model", "type": "llm", "publisher": "lmstudio-community", "arch": "qwen2", "compatibility_type": "gguf", "quantization": "Q4_K_M", "state": "not-loaded", "max_context_length": 131072 }, { "id": "pixtral-12b", "object": "model", "type": "vlm", "publisher": "mlx-community", "arch": "pixtral", "compatibility_type": "mlx", "quantization": "8bit", "state": "not-loaded", "max_context_length": 1024000 }, { "id": "mistral-nemo-instruct-2407", "object": "model", "type": "llm", "publisher": "unsloth", "arch": "llama", "compatibility_type": "gguf", "quantization": "F16", "state": "not-loaded", "max_context_length": 1024000 }, { "id": "qwen2.5-coder-32b-instruct-128k", "object": "model", "type": "llm", "publisher": "unsloth", "arch": "qwen2", "compatibility_type": "gguf", "quantization": "Q4_K_M", "state": "not-loaded", "max_context_length": 131072 }, { "id": "qwen2.5-coder-32b-instruct", "object": "model", "type": "llm", "publisher": "Qwen", "arch": "qwen2", "compatibility_type": "gguf", "quantization": "Q4_0", "state": "not-loaded", "max_context_length": 131072 } ], "object": "list" }

[teslanaut]

Here is a video of it in action. I have LM Studio running but not Ollama on purpose.

recording.mp4

[notpeter]

Got it. I was able to reproduce with zelensky-78b. Thanks @teslanaut

    {
      "id": "zelensky-78b",
      "object": "model",
      "type": "llm",
      "publisher": "nisten",
      "arch": "qwen2",
      "compatibility_type": "gguf",
      "state": "loaded",
      "max_context_length": 98304,
      "loaded_context_length": 4096
    }

[notpeter]

Fixed in Zed Stable v0.171.5 / Zed Preview v0.171.5-pre.