mlx-optiq
← Models

NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16, quantized by OptiQ for Apple Silicon; takes images.

vision
On disk7.5 GB
Memoryabout 10 GB
Parent score68.92 gemma-4-e4b-it-OptiQ-4bit
Imagesyes

This model reuses the per-layer bit allocation of gemma-4-e4b-it-OptiQ-4bit, whose measured Capability Score is 68.92.

The parent's Capability Score breakdown

This model has no Capability Score of its own. The chart shows the six metrics behind its parent gemma-4-e4b-it-OptiQ-4bit's score of 68.92, taken from the parent's model card and compared with uniform 4-bit.

  • gemma-4-e4b-it-OptiQ-4bit
  • Uniform 4-bit
The parent's Capability Score breakdown: gemma-4-e4b-it-OptiQ-4bitMMLU: gemma-4-e4b-it-OptiQ-4bit 58.8, Uniform 4-bit 52.9; GSM8K: gemma-4-e4b-it-OptiQ-4bit 77.8, Uniform 4-bit 46.1; IFEval: gemma-4-e4b-it-OptiQ-4bit 70.6, Uniform 4-bit 68.6; BFCL-V3: gemma-4-e4b-it-OptiQ-4bit 87.5, Uniform 4-bit 85.0; HumanEval: gemma-4-e4b-it-OptiQ-4bit 76.8, Uniform 4-bit 58.5; HashHop: gemma-4-e4b-it-OptiQ-4bit 42.0, Uniform 4-bit 20.0MMLU58.852.9GSM8K77.846.1IFEval70.668.6BFCL-V387.585.0HumanEval76.858.5HashHop42.020.0

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/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

OptiQ Code

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

optiq serve --model mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit
optiq code

OptiQ Lab

optiq lab --model mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

Claude Code

optiq run claude -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

Codex

optiq run codex -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

OpenCode

optiq run opencode -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

Mistral Vibe

optiq run vibe -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

Hermes Agent

optiq run hermes -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

OpenClaw

optiq run openclaw -m mlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit

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/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilyGemma-4 · Gemma-4 docs
ArchitectureGemma 4
Base modelnvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16
Size on disk7.5 GB
Memory neededabout 10 GB
Imagesyes, through optiq serve and the Lab
Family referenceThis model reuses the per-layer bit allocation of gemma-4-e4b-it-OptiQ-4bit, whose measured Capability Score is 68.92.
Downloads61 in the last month
Updated2026-10-24
Hugging Facemlx-community/NVIDIA-Nemotron-3.5-Bolt-E4B-OptiQ-4bit