Project Files
benchmark / live-results / gemma-4-26b-a4b-mega-01.md
BLOCKED — model failed to load.
LM Studio error: Failed to load model: ValueError: The mlx-vlm batched vision path does not support KV cache quantization yet.
This is a real environment/framework limitation (MLX vision-model batching doesn't support the KV cache quantization mode configured), not something fixable from the chat UI without changing model load parameters — which the user explicitly instructed not to touch. Not retried with different load settings for that reason.
Verdict: BLOCKED, not run.