Skip to content
Navigation menu
Search
Powered by Algolia
Search
Log in
Create account
DEV Community
Close
Edge AI in Practice Series' Articles
Back to GINIGEN AI's Series
Running a 180B Model on a Laptop With No GPU: How 4-bit GGUF Keeps Full Accuracy
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 6
Running a 180B Model on a Laptop With No GPU: How 4-bit GGUF Keeps Full Accuracy
#
ai
#
llm
#
edgeai
#
quantization
Comments
Add Comment
6 min read
Moderation
Building an AI That Refuses to Say "Safe": Reliability Engineering for an Offline Disaster App
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 6
Building an AI That Refuses to Say "Safe": Reliability Engineering for an Offline Disaster App
#
ai
#
edgeai
#
llm
#
android
Comments
Add Comment
7 min read
Moderation
4-bit GGUF Quality for MoE Models: Why Only 3B of 180B Params Fire, and How to Prove Parity
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 6
4-bit GGUF Quality for MoE Models: Why Only 3B of 180B Params Fire, and How to Prove Parity
#
ai
#
llm
#
quantization
#
edgeai
Comments
Add Comment
7 min read
Moderation
One Self-Improving Model, Eleven Number-One Titles: What That Takes
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 11
One Self-Improving Model, Eleven Number-One Titles: What That Takes
#
ai
#
machinelearning
#
llm
#
opensource
Comments
1
comment
2 min read
Moderation
Zero-Token Decisions in the Browser: On-Device Multimodal Classification with EmbeddingGemma 2 and WebGPU
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 12
Zero-Token Decisions in the Browser: On-Device Multimodal Classification with EmbeddingGemma 2 and WebGPU
#
ai
#
machinelearning
#
webgpu
#
opensource
Comments
Add Comment
3 min read
Moderation
LFM2.5-24B-A2B on Your Own Hardware: The RAM, KV Cache and Bandwidth Budget
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 19
LFM2.5-24B-A2B on Your Own Hardware: The RAM, KV Cache and Bandwidth Budget
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
11 min read
Moderation
0.1 Bits Per Weight Is Not 0.1 Bits Per Weight: Where the Bytes Go in Sub-2-Bit LLMs
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 21
0.1 Bits Per Weight Is Not 0.1 Bits Per Weight: Where the Bytes Go in Sub-2-Bit LLMs
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
14 min read
Moderation
The Inference Hardware You Already Own Might Train Models Too: A 325M Pretraining Run on an NPU
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 26
The Inference Hardware You Already Own Might Train Models Too: A 325M Pretraining Run on an NPU
#
ai
#
machinelearning
#
edgeai
#
hardware
Comments
Add Comment
3 min read
Moderation
From 46M to 325M: An Inference NPU Now Pretrains a Language Model at GPU Quality
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 29
From 46M to 325M: An Inference NPU Now Pretrains a Language Model at GPU Quality
#
ai
#
machinelearning
#
edgeai
#
npu
Comments
Add Comment
2 min read
Moderation
LFM3-1.7B On-Device: The 256K KV Cache Math, Which GGUF to Ship, and the Safety Test to Run First
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Oct 31
LFM3-1.7B On-Device: The 256K KV Cache Math, Which GGUF to Ship, and the Safety Test to Run First
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
14 min read
Moderation
Why a Sparse 35B MoE Runs Faster on a CPU Than a Dense 27B Model
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 1
Why a Sparse 35B MoE Runs Faster on a CPU Than a Dense 27B Model
#
ai
#
machinelearning
#
edgeai
#
llm
Comments
Add Comment
5 min read
Moderation
Asymmetric Control for AI Agents: Why the Stop Path Must Be Independent of the Model
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 1
Asymmetric Control for AI Agents: Why the Stop Path Must Be Independent of the Model
#
ai
#
security
#
agents
#
architecture
Comments
Add Comment
3 min read
Moderation
A 763B MoE at "2 Bits": Where the Bytes Really Go in the GSQ-RCO GGUFs, and Why 100% Recovery Sits Inside the Noise
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 2
A 763B MoE at "2 Bits": Where the Bytes Really Go in the GSQ-RCO GGUFs, and Why 100% Recovery Sits Inside the Noise
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
15 min read
Moderation
A 763B MoE on a 12 GB GPU: Our Speed Ceiling Versus What Strata Reports, and What Closes the Gap
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 3
A 763B MoE on a 12 GB GPU: Our Speed Ceiling Versus What Strata Reports, and What Closes the Gap
#
ai
#
machinelearning
#
edgeai
#
llm
Comments
Add Comment
13 min read
Moderation
Kimi K3 in 8 GB of RAM: Why 28x More Memory Buys Only 1.7x Speed, and the Byte Math That Predicts It
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 5
Kimi K3 in 8 GB of RAM: Why 28x More Memory Buys Only 1.7x Speed, and the Byte Math That Predicts It
#
ai
#
opensource
#
edgeai
#
llm
Comments
Add Comment
12 min read
Moderation
Does 2-Bit Quantization Break On-Device Safety? What an EMNLP 2026 Study Found, and a GGUF Release Gate
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 7
Does 2-Bit Quantization Break On-Device Safety? What an EMNLP 2026 Study Found, and a GGUF Release Gate
#
ai
#
machinelearning
#
edgeai
#
security
1
reaction
Comments
Add Comment
14 min read
Moderation
One Omni Model, Two GGUF Files: What LFM3-Omni-2B Really Costs in RAM, from the Tensor Headers Up
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 8
One Omni Model, Two GGUF Files: What LFM3-Omni-2B Really Costs in RAM, from the Tensor Headers Up
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
12 min read
Moderation
Cutting 64 of 256 Experts from a GGUF by Byte Copy: What the Headers Say About GLM-5.3-SLIM on One B200
GINIGEN AI
GINIGEN AI
GINIGEN AI
Follow
Nov 11
Cutting 64 of 256 Experts from a GGUF by Byte Copy: What the Headers Say About GLM-5.3-SLIM on One B200
#
ai
#
machinelearning
#
edgeai
#
quantization
Comments
Add Comment
13 min read
Moderation
We're a place where coders share, stay up-to-date and grow their careers.
Log in
Create account