Configuration Parsing Warning:Config file tokenizer_config.json cannot be fetched (too big)

ACE-Step MLX Planner 0.6B FP16

MLX-compatible fp16 package for the ACE-Step 5Hz planner LM used by optional Phase 6 LM-conditioning. This is the recommended planner for low-memory device runs.

The planner is shared by the base turbo and XL-turbo ACE-Step MLX model packages so it is not duplicated in every DiT repo. The 5Hz planner ships in three sizes (0.6B / 1.7B / 4B); the host app selects a size by available memory.

Usage

Baseline generation does not need this repo. To enable Phase 6, pass the planner snapshot to the generator with --lm-weights. For Q4 DiT variants, also pass the model repo's bundled detokenizer.safetensors as --lm-conditioning-checkpoint.

Source attribution

The planner LM is the ACE-Step 5Hz planner, a fine-tune in the Qwen3 family (Apache 2.0). This package changes the on-disk layout.

NOTICE

This package carries the upstream Qwen3 NOTICE obligation. See the upstream Qwen3 model card for the canonical NOTICE text.

Downloads last month: 23

Safetensors

Model size

0.7B params

Tensor type

BF16

MLX

Hardware compatibility

Quantized

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support