Search results

VisionModelWrapper: single-pass vision prefill causes large VRAM spike on long contexts

Status: Open.

#289 In lmstudio-ai/mlx-engine;

· AirRunner opened on Mar 17, 2026

VisionModelKit._reset_for_prediction() calls _full_model_init() on every request

Status: Open.

#288 In lmstudio-ai/mlx-engine;

· AirRunner opened on Mar 17, 2026

Cross-turn KV cache for VisionModelKit sequential generation

Status: Open.

#287 In lmstudio-ai/mlx-engine;

· AirRunner opened on Mar 17, 2026

MLX vision models do not currently support KV cache quantization for qwen3.5

Status: Open.

#286 In lmstudio-ai/mlx-engine;

· kay1ess opened on Mar 7, 2026

Wide performance discrepancy between LM Studio MLX and mlx_vlm.generate with Qwen3.5-35B-A3B

Status: Open.

#285 In lmstudio-ai/mlx-engine;

· Belluxx opened on Mar 6, 2026

Error when loading model: ValueError: Model type qwen3_5 not supported.

Status: Open.

#284 In lmstudio-ai/mlx-engine;

· wqw547243068 opened on Mar 3, 2026

QK-Normalization Support

Status: Open.

#283 In lmstudio-ai/mlx-engine;

· JakobStadlhuber opened on Feb 28, 2026

mode was ignored in model_kit.vision_add_ons.load_utils.maybe_apply_quantization, causing mxfp8 and such quants to fail.

Status: Open.

#282 In lmstudio-ai/mlx-engine;

· Foxtr0t1337 opened on Feb 22, 2026

Speculative decoding fails (almost) silently instead of clearly notifying users about draft model loading issues

Status: Open.

#280 In lmstudio-ai/mlx-engine;

· vnicolici opened on Feb 20, 2026

Support for ouro Architecture

Status: Open.

#277 In lmstudio-ai/mlx-engine;

· Flynni123 opened on Feb 16, 2026

🥲 Failed to load the model:- lfm2-vl-3b

Status: Open.

#275 In lmstudio-ai/mlx-engine;

· shahrukh-alizai opened on Feb 15, 2026

SpeculativeDecodingNotSupportedError: Speculative decoding is not supported for batched MLX models.

Status: Open.

#269 In lmstudio-ai/mlx-engine;

· Clete2 opened on Feb 7, 2026