mlx-optiq
← Models

Spark-X2.6-4B-OptiQ-4bit

XHToken/Spark-X2.6-4B, quantized by OptiQ for Apple Silicon.

On disk3.0 GB
Memoryabout 5 GB
Parent score64.69 Spark-X2.5-4B-OptiQ-4bit

This model reuses the per-layer bit allocation of Spark-X2.5-4B-OptiQ-4bit, whose measured Capability Score is 64.69.

The parent's Capability Score breakdown

This model has no Capability Score of its own. The chart shows the six metrics behind its parent Spark-X2.5-4B-OptiQ-4bit's score of 64.69, taken from the parent's model card and compared with uniform 4-bit.

  • Spark-X2.5-4B-OptiQ-4bit
  • Uniform 4-bit
The parent's Capability Score breakdown: Spark-X2.5-4B-OptiQ-4bitMMLU: Spark-X2.5-4B-OptiQ-4bit 66.8, Uniform 4-bit 63.1; GSM8K: Spark-X2.5-4B-OptiQ-4bit 87.0, Uniform 4-bit 84.9; IFEval: Spark-X2.5-4B-OptiQ-4bit 75.2, Uniform 4-bit 75.8; BFCL-V3: Spark-X2.5-4B-OptiQ-4bit 75.5, Uniform 4-bit 70.0; HumanEval: Spark-X2.5-4B-OptiQ-4bit 75.6, Uniform 4-bit 74.4; HashHop: Spark-X2.5-4B-OptiQ-4bit 8.0, Uniform 4-bit 1.0MMLU66.863.1GSM8K87.084.9IFEval75.275.8BFCL-V375.570.0HumanEval75.674.4HashHop8.01.0

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/Spark-X2.6-4B-OptiQ-4bit

OptiQ Code

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

optiq serve --model mlx-community/Spark-X2.6-4B-OptiQ-4bit
optiq code

OptiQ Lab

optiq lab --model mlx-community/Spark-X2.6-4B-OptiQ-4bit

Claude Code

optiq run claude -m mlx-community/Spark-X2.6-4B-OptiQ-4bit

Codex

optiq run codex -m mlx-community/Spark-X2.6-4B-OptiQ-4bit

OpenCode

optiq run opencode -m mlx-community/Spark-X2.6-4B-OptiQ-4bit

Mistral Vibe

optiq run vibe -m mlx-community/Spark-X2.6-4B-OptiQ-4bit

Hermes Agent

optiq run hermes -m mlx-community/Spark-X2.6-4B-OptiQ-4bit

OpenClaw

optiq run openclaw -m mlx-community/Spark-X2.6-4B-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/Spark-X2.6-4B-OptiQ-4bit")
print(generate(model, tokenizer, prompt="Hello", max_tokens=256))

Details

FamilySpark-X2.5 · Spark-X2.5 docs
ArchitectureSpark-X2.5
Base modelXHToken/Spark-X2.6-4B
Size on disk3.0 GB
Memory neededabout 5 GB
Imagesno, text only
Family referenceThis model reuses the per-layer bit allocation of Spark-X2.5-4B-OptiQ-4bit, whose measured Capability Score is 64.69.
Downloads0 in the last month
Updated2026-10-24
Hugging Facemlx-community/Spark-X2.6-4B-OptiQ-4bit