gemma-4-E4B-RotorQuant-MLX-2bit

[!TIP] KV-cache quantization without any fork (recommended, 2026): upstream llama.cpp/Ollama now cover this natively — use -ctk q8_0 -ctv q8_0 (~half KV memory, negligible quality loss: perplexity +0.002–0.05) or -ctk q4_0 -ctv q4_0 (~quarter memory, ≈7.6% perplexity increase). In Ollama: OLLAMA_KV_CACHE_TYPE=q8_0 with OLLAMA_FLASH_ATTENTION=1. Keep K and V types symmetric to stay on the fast fused Flash-Attention path. Since April 2026, mainline llama.cpp also applies Hadamard rotation to KV activations (PR #21038), which greatly improves low-bit KV quality (opt-out: LLAMA_ATTN_ROT_DISABLE=1).

The RotorQuant/TurboQuant fork flow below is experimental/legacy: the TurboQuant llama.cpp PR was closed without merging (June 2026) and the fork is unmaintained relative to mainline. It is NOT required to use this model.

<!-- kv-upstream-note -->

Gemma 4 E4B - RotorQuant MLX 2-bit

2-bit weight-quantized MLX version of google/gemma-4-E4B with the legacy RotorQuant KV-cache fork (superseded by upstream llama.cpp KV options). Optimized for Apple Silicon inference via the MLX framework. The most aggressive quantization, fitting the full model in the smallest possible footprint.

Approximate model size: ~1.2 GB

Model Specifications

Quickstart

import mlx.core as mx
from mlx_lm import load, generate

model, tokenizer = load("majentik/gemma-4-E4B-RotorQuant-MLX-2bit")

prompt = "The history of artificial intelligence began"
response = generate(model, tokenizer, prompt=prompt, max_tokens=512)
print(response)

For multimodal usage with images:

from mlx_vlm import load, generate

model, processor = load("majentik/gemma-4-E4B-RotorQuant-MLX-2bit")

prompt = "Describe the contents of this image."
output = generate(model, processor, prompt=prompt, image="path/to/image.jpg", max_tokens=512)
print(output)

About the RotorQuant / TurboQuant labels

RotorQuant and TurboQuant are this project's release labels, not distinct quantization algorithms — for any given tier, both brand repos carry byte-identical weights produced with the standard MLX / llama.cpp quantizers. No brand-specific speedup is claimed or measured. The KV-cache fork these labels originally referred to is legacy; for KV-cache memory savings use the upstream options described above (-ctk/-ctv q8_0, OLLAMA_KV_CACHE_TYPE).

KV-Cache Quantization Comparison

Memory Estimates (Gemma 4 E4B)

Hardware Requirements

This model requires approximately 1.2 GB of unified memory. Recommended hardware:

• Apple M1 (8 GB+)
• Apple M2 (8 GB+)
• Apple M3 (8 GB+)
• Apple M4 (8 GB+)
• Any Apple Silicon Mac with 8 GB+ unified memory

See Also

• google/gemma-4-E4B -- Base model
• majentik/gemma-4-E4B-RotorQuant-MLX-8bit -- MLX 8-bit variant
• majentik/gemma-4-E4B-RotorQuant-MLX-4bit -- MLX 4-bit variant
• RotorQuant GitHub
• MLX Framework

Quant trade-off (MLX lane)

(Current variant — 2bit — is bolded.)

Variants in this family

(Showing 14 sibling variants under majentik/gemma-4-e4b-*. The current variant — RotorQuant-MLX-2bit — is bolded.)