mudler/LFM2-24B-A2B-APEX-GGUF

🚨⚠️ I HAVE REACHED HUGGING FACE'S FREE STORAGE LIMIT ⚠️🚨

Uploading new APEX quants is getting harder — beyond HF storage limits, I also no longer have access to a beefy GPU to quantize larger MoE models.
I host 25+ free APEX MoE quantizations as an independent contributor and this work is unpaid.
I'll keep doing my best to publish new models — your support makes it possible to continue.

🎉 Patreon (Monthly)  |  ☕ Buy Me a Coffee  |  ⭐ GitHub Sponsors

Every contribution goes directly toward Hugging Face storage fees and compute to keep APEX quants free for everyone.

LFM2-24B-A2B APEX GGUF

APEX (Adaptive Precision for EXpert Models) quantizations of LFM2-24B-A2B by LiquidAI.

Brought to you by the LocalAI team | APEX Project | Technical Report

Benchmark Results

Benchmarks coming soon. For reference APEX benchmarks on the Qwen3.5-35B-A3B architecture, see mudler/Qwen3.5-35B-A3B-APEX-GGUF.

What is APEX?

APEX is a quantization strategy for Mixture-of-Experts (MoE) models. It classifies tensors by role (routed expert, shared expert, attention) and applies a layer-wise precision gradient -- edge layers get higher precision, middle layers get more aggressive compression. I-variants use diverse imatrix calibration (chat, code, reasoning, tool-calling, agentic traces, Wikipedia).

See the APEX project for full details, technical report, and scripts.

Architecture

• Model: LFM2-24B-A2B (lfm2_moe) by LiquidAI
• Layers: 40 (30 convolutional + 10 full attention, hybrid)
• Experts: 64 routed (4 active per token) + 2 dense layers
• Total Parameters: 24B
• Active Parameters: ~2B per token
• APEX Config: 5+5 symmetric edge gradient across 40 layers
• Calibration: v1.3 diverse dataset (chat, code, reasoning, multilingual, tool-calling, Wikipedia)

Run with LocalAI

local-ai run mudler/LFM2-24B-A2B-APEX-GGUF@LFM2-24B-A2B-APEX-I-Balanced.gguf

Credits

APEX is brought to you by the LocalAI team. Developed through human-driven, AI-assisted research. Built on llama.cpp.