failures running many GGUF embedding models

[arvindvenkataramani]

Which version of LM Studio?
Example: LM Studio 0.4.6

Which operating system?
MacOS Sequoia 15.6.4

What is the bug?
I get errors with GGUF versions of various embedding models when trying to use them with knowledge stacks in Msty Studio.

Error claims out of memory, but I have a 48GB M4 Max and LMS says only 10 GB is used (verified with Activity Monitor)

I am able to successfully use text-embedding-qwen3-embedding-0.6b, text-embedding-nomic-embed-text-v1.5@q8_0, and text-embedding-mxbai-embed-large-v1

Logs
Attempting to use the Qwen 3 4b and 8b embedding models:

2026-03-13 23:13:24 [DEBUG]
 [lmstudio-llama-cpp] LLM: Embedding failed: Failed during string embedding. Message: Unknown exception caused embedding to stop: Failed to decode batch! Error:
2026-03-13 23:13:24 [DEBUG]
 [WARNING] At least one last token in strings embedded is not SEP. 'tokenizer.ggml.add_eos_token' should be set to 'true' in the GGUF header
2026-03-13 23:13:24 [DEBUG]
 ggml_metal_synchronize: error: command buffer 0 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_synchronize: error: command buffer 0 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
2026-03-13 23:13:24 [DEBUG]
 ggml_metal_synchronize: error: command buffer 0 failed with status 5
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
      [repeated several times]
error: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory)
ggml_metal_graph_compute: backend is in error state from a previous command buffer failure - recreate the backend to recover
graph_compute: ggml_backend_sched_graph_compute_async failed with error -1
process_ubatch: failed to compute graph, compute status: -1
decode: removing memory module entries for seq_id = 0, pos = [0, +inf)
llama_decode: failed to decode, ret = -3
BatchDecode : failed to decode

llama-embed-nemotron-8b (GGUF) and the model is not auto-loading on request, and when manually loaded fails with

 No models loaded. Please load a model in the developer page or use the 'lms load' command.```

**To Reproduce**
Steps to reproduce the behavior:
1. install GGUF embedding models
2. Select embedding models in MstyStudio
3. Attempt to compose embeddings
4. Fails with one of these two categories of errors