Recipe

gemma-4-26b-a4b-it-qat-4bit-apple-m5-max-128gb-mlxfast-tp1

gemma-4-26b-a4b-it-qat-4bit-apple-m5-max-128gb-mlxfast-tp1

Official mlx.fast Gemma 4 26B A4B cohort score on the organizer M5 Max 128GB box. Candidate evidence, not a digest-pinned launch contract.

Record

Status: candidate
Source: mlxfast
Engine: mlx
Engine version: gemma4-26b-a4b-mlx-v1
Accelerators: 1
Tensor parallel: 1
Context tokens: 262,144
Max concurrency: 8
chat: unknown
reasoning: unknown
tools: unknown
vision: unknown

Hugging Face model card

Identity

https://huggingface.co/mlx-community/gemma-4-26B-A4B-it-qat-4bit

Repository: mlx-community/gemma-4-26B-A4B-it-qat-4bit
Status: known
Link type: Exact Hub repository

Public Hugging Face repository confirmed by the Hub API.

Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.

Observed configuration

mlx

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source: https://www.yukon.org/mlxfast

Observed source tokens

Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.

1. git
2. clone
3. https://github.com/Layr-Labs/mlxfast-gemma4-26b-a4b-engine
1. ./tools/fetch-benchd.sh
1. ./setup.sh
1. ./setup-gemma4-assistant.sh

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status	Sweep
8	—	6,950.7	484.5	—	observed	gemma-4-26b-a4b-it-qat-4bit-apple-m5-max-128gb-mlxfast-tp1-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count: 1
hardware id: apple-m5-max-128gb
id: gemma-4-26b-a4b-it-qat-4bit-apple-m5-max-128gb-mlxfast-tp1
model instance id: mlx-community-gemma-4-26b-a4b-it-qat-4bit--4bit
recipe source: mlxfast
schema version: local-ai-registry/v1
speed sweep ids: gemma-4-26b-a4b-it-qat-4bit-apple-m5-max-128gb-mlxfast-tp1-sweep
status: candidate

capabilities

engine

name: mlx
version: gemma4-26b-a4b-mlx-v1

serving

max concurrency: 8
max context tokens: 262,144
tensor parallel: 1

Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at: 2026-08-30T09:28:11Z

capabilities.chat · reason capability-not-verified

capabilities.chat · state unknown

capabilities.reasoning · provenance · captured at: 2026-08-30T09:28:11Z

capabilities.reasoning · reason capability-not-verified

capabilities.reasoning · state unknown

capabilities.tools · provenance · captured at: 2026-08-30T09:28:11Z

capabilities.tools · reason capability-not-verified

capabilities.tools · state unknown

capabilities.vision · provenance · captured at: 2026-08-30T09:28:11Z

capabilities.vision · reason capability-not-verified

capabilities.vision · state unknown

engine.graph mode · provenance · captured at: 2026-08-30T09:28:11Z

engine.graph mode · reason runtime-detail-not-published

engine.graph mode · state unknown

serving.kv cache tokens · provenance · captured at: 2026-08-30T09:28:11Z

serving.kv cache tokens · reason kv-cache-capacity-not-published

serving.kv cache tokens · state unknown

metadata

mlxfast · engine repository: github.com/Layr-Labs/mlxfast-gemma4-26b-a4b-engine ↗
mlxfast · leaderboard: www.yukon.org/mlxfast ↗
mlxfast · observed command: git clone https://github.com/Layr-Labs/mlxfast-gemma4-26b-a4b-engine && ./tools/fetch-benchd.sh && ./setup.sh && ./setup-gemma4-assistant.sh
mlxfast · official hardware: organizer M5 Max 128GB
mlxfast · solver: samfenwick
mlxfast · tokenized · arguments: ./setup-gemma4-assistant.sh
mlxfast · tokenized · fidelity: faithful
mlxfast · track id: gemma4-26b-a4b-mlx-v1

mlxfast · tokenized · steps

0	1	2
git	clone	github.com/Layr-Labs/mlxfast-gemma4-26b-a4b-engine ↗
./tools/fetch-benchd.sh	Unknown	Unknown
./setup.sh	Unknown	Unknown
./setup-gemma4-assistant.sh	Unknown	Unknown

provenance

captured at: 2026-08-30T09:28:11Z

sources

captured at	kind	url
2026-08-30T09:28:11Z	normalized-recipe	www.yukon.org/mlxfast ↗