Search results

Gemma 4 model family: three independent loader gaps blocking Apple Silicon adoption

Status: Open.

#282 In ml-explore/mlx-swift-lm;

· swinxx opened on May 12, 2026

Register gemma4_assistant model type so Gemma 4 MTP drafters load through LLMModelFactory

Status: Open.

#279 In ml-explore/mlx-swift-lm;

· brdcastro opened on May 9, 2026

[BUG] TokenLoopHandler should not require Sendable

Status: Open.

#272 In ml-explore/mlx-swift-lm;

· davidkoski opened on May 7, 2026

support gpt-oss-20b-tq3

Status: Open.

#270 In ml-explore/mlx-swift-lm;

· mutou981 opened on May 7, 2026

Add convert API for in-Swift model quantization

Status: Open.

#266 In ml-explore/mlx-swift-lm;

· programVeins opened on May 4, 2026

[BUG] Crashes when trying to load quantized model (e.g: Gemma4 MXFP4)

Status: Open.

#261 In ml-explore/mlx-swift-lm;

· coreintelligence opened on May 2, 2026

Gemma 4: tool calls never extracted (ToolCallFormat.infer misses "gemma4" + GemmaFunctionParser uses outdated tags)

Status: Open.

#259 In ml-explore/mlx-swift-lm;

· angelsbrood opened on May 1, 2026

Gemma 4 26B-A4B 8-bit: broadcast crash on first generate when repetitionPenalty is set

Status: Open.

#258 In ml-explore/mlx-swift-lm;

· angelsbrood opened on May 1, 2026

[BUG] Gemma 4 VLM: crash on continuation request with manual KV cache reuse (1D input shape)

Status: Open.

#240 In ml-explore/mlx-swift-lm;

· brdcastro opened on Apr 25, 2026

Gemma4 (MLXLLM): K-eq-V attention path double-transposes V, causes broadcast_shapes crash

Status: Open.

#231 In ml-explore/mlx-swift-lm;

· dylandeheer opened on Apr 22, 2026

Upstreaming improvements from fast-moving forks

Status: Open.

#221 In ml-explore/mlx-swift-lm;

· ronaldmannak opened on Apr 17, 2026

TokenRing.loadPrompt crash: prompt.dim(0) returns batch dim for 2D tensors

Status: Open.

#220 In ml-explore/mlx-swift-lm;

· VincentGourbin opened on Apr 17, 2026