mlx-optiq
← Models

gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit, quantized by OptiQ for Apple Silicon; mixture-of-experts, takes images.

visionMoEexpert-pruned
On disk11.2 GB
Memoryabout 14 GB
Capability Score68.13
Imagesyes

Capability Score breakdown

The six metrics behind this model's Capability Score of 68.13, as reported in the model card. Values are compared with unpruned parent.

  • This model
  • Unpruned parent
Capability Score breakdown: gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14BMMLU: This model 38.9, Unpruned parent 65.0; GSM8K: This model 89.1, Unpruned parent 93.8; IFEval: This model 80.8, Unpruned parent 73.0; BFCL-V3: This model 90.0, Unpruned parent 91.5; HumanEval: This model 72.0, Unpruned parent 90.2; HashHop: This model 38.0, Unpruned parent 41.0MMLU38.965.0GSM8K89.193.8IFEval80.873.0BFCL-V390.091.5HumanEval72.090.2HashHop38.041.0

Use it

Install

pip install mlx-optiq

Serve it

An OpenAI- and Anthropic-compatible server on this Mac. Every app below can use it.

optiq serve --model mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

OptiQ Code

The terminal coding agent uses whatever model optiq serve is running.

optiq serve --model mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B
optiq code

OptiQ Lab

optiq lab --model mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

Claude Code

optiq run claude -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

Codex

optiq run codex -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

OpenCode

optiq run opencode -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

Mistral Vibe

optiq run vibe -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

Hermes Agent

optiq run hermes -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

OpenClaw

optiq run openclaw -m mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B

optiq run starts optiq serve if it is not running, sets the agent up for this session only, and stops the server when the agent exits.

mlx-lm, text only

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilyGemma-4 · Gemma-4 docs
ArchitectureGemma 4
Base modelmlx-community/gemma-4-26B-A4B-it-OptiQ-4bit
Size on disk11.2 GB
Memory neededabout 14 GB
Imagesyes, through optiq serve and the Lab
Capability Score68.13 (from the model card)
Downloads695 in the last month
Updated2026-10-18
Hugging Facemlx-community/gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B