Local AI · Recipe · 1× Mac

GLM-5.3-Flash 320B-A18B Q4_K with mlx-serve (ds4) on 1× Mac (M5 Ultra)

GLM-5.3-Flash 320B-A18B Q4_K with mlx-serve (ds4) on 1× Mac (M5 Ultra): 35.8 tok/s according to github.com (dataset as of Sep 29, 2026).

by drowzeys

35.8tok/sEveryday

1 request, llmprobe standard prompt, thinking on, with embedded MTP (2 draft tokens) (per recipe)

Source: github.com

Intelligence (original model) · Artificial Analysis42with thinking

Instructions in the README Repo

Engine: mlx-serve 26.9.7-dev @933566c + ds4 9139e2a (GGUF-Backend)
Quantization: GGUF Q4_K (antirez, 177.8 GiB), kompakter DSA-KV
Model family: GLM-5.3-Flash
Context: 262,144
Parameters: 320B MoE / 18B active
Creator: drowzeys
GitHub stars: 0
Repo updated: Sep 30, 2026

Measurements

Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.

Decode · 1 request

Everyday (decode)35.8 tok/sLlmprobe 0.6.13, greedy, Median aus 3 (35.6–37.5), 192 Token Source
Peak (decode)41 tok/sVorhersagbarer Prompt (Spekulation, 1.96 tok/step) Source

Prefill by context

9.4K439 tok/s Source

Time to first token

Shorter is better.

Prompt not stated427 mscontext kurz Source

What you need

Hardware: 1 × Mac / Apple Silicon
Weights: Main weightsantirez/glm-5.3-flash-gguf License unclear
Basezai-org/GLM-5.3-Flash MIT
Engine: mlx-serve 26.9.7-dev @933566c + ds4 9139e2a (GGUF-Backend)
Context: 262,144 tokens

Notes

What matters before you rebuild it.

Experimental
No license on HF
License unclear

Sources

Related recipes

4× DGX Spark

GLM-5.3-Flash NVFP4 with SparkRing on 4× DGX Spark

by fujitsupolycom

62.1tok/sEveryday

1 request, decode at 16K context (profile table), thinking always, max Source

Intelligence42with thinking

Custom kernel required

NVFP4 Spark (Local Inference Lab)vLLM (local-inference-lab Fork) via SparkRing (switchless NCCL ueber ConnectX-7)Weights Repo Updated Oct 6, 2026

Details

2× DGX Spark

GLM-5.3-Flash EXL3 with TensorFold on 2× DGX Spark (1M context, 4 streams)

by MiaAI-Lab

60.4tok/sEveryday

Prose decode, 1 request; quality (FP8 KV, 4-bit dense): GSM8K 98.8% / HumanEval 95.7% / HumanEval+ & MBPP+ 86.5% Source

Intelligence42with thinking

≤ 3 bit: quantization may cost quality
Custom kernel required
Non-commercial
License unclear

EXL3/TR3 4bpw routed experts + 4-bit dense, BF16 elsewhere, FP8 KVTensorFold v0.6.0 + Recipe v1.10 (README: 96 Patches; gepinntes Image v0.6.0-e66c657a433d mit 107 Patches: DFlash2+Copy-Drafts, 4-bit dense, FP8-KV, RoCE-One-Shot-All-Gather, Shared KV-Pool, optionaler Spill-Tier, Vision, Tool-Calling, Anthropic /v1/messages; PARALLEL=8 opt-in)Weights Repo Updated Oct 10, 2026

Details

4× DGX Spark

GLM-5.3-Flash 320B-A18B NVFP4 with vLLM on 4× DGX Spark

by tonyd2wild

40.8tok/sEveryday

1 request, prose prompt, with DFlash2 (per recipe) Source

Intelligence42with thinking

Custom kernel required
Non-commercial

NVFP4 weights + FP8 KV (Lane B: NVFP4 KV)vLLM 0.28.0Weights Repo Updated Oct 3, 2026

Details

3× DGX Spark

GLM-5.3-Flash 320B EXL3 4 bpw with vLLM on 3× DGX Spark

by MiaAI-Lab

39.6tok/sEveryday

1 request, prose prompt, with DFlash2 (per recipe) Source

Intelligence42with thinking

≤ 3 bit: quantization may cost quality
Custom kernel required
Non-commercial
License unclear

EXL3/TR3 4 bpw (routed experts), FP8 KV (fp8_ds_mla)vLLMWeights Repo Updated Oct 4, 2026

Details

2× DGX Spark

GLM-5.3-Flash 320B-A18B EXL3 4 bpw with ExLlamaV3 on 2× DGX Spark

by MiaAI-Lab

36.1tok/sEveryday

1 request, prose prompt, with DFlash2 (per recipe) Source

Intelligence42with thinking

≤ 3 bit: quantization may cost quality
Custom kernel required
Non-commercial
License unclear

EXL3 4bpw (TR3) + FP8 KV (fp8_ds_mla)TensorFold-basiertes Kit (ghcr.io/miaai-lab/glm-5.3-flash-2x-dgx-sparks:exl3-instanttensor), DFlash2 7 draft tokens; opt-in Speed-Boost-Profil (einmaliger ~25-min-Build auf dem Head)Weights Repo Updated Oct 4, 2026

Details

2× DGX Spark