llm-kb
← Back to social
Community

Community member extends Gemma 4 to 40.5B parameters via layer insertion

A Reddit user, TOTORONG, successfully extended Google's Gemma 4 31B model to 40.5B parameters by inserting additional layers, releasing the result as extGemma4-40_5B on Hugging Face. The project builds on previous attempts documented in earlier posts, overcoming challenges where new inserted layers initially degraded performance. This demonstrates community-driven model customization and fine-tuning techniques.

31 engagement·1 source·Sun, Jul 12, 2026, 03:49 AM
TOTORONG shared on Reddit that they have returned to the project of extending Gemma 4, a model originally released by Google with 31 billion parameters. The new model, extGemma4-40_5B, adds approximately 9.5B parameters through layer insertion, bringing the total to 40.5B. The user referenced two previous posts detailing the initial extension to 44B (88 layers) and subsequent issues where inserted layers killed performance. The Hugging Face repository is available at https://huggingface.co/TOTORONG/extGemma4-40_5B. The post is AI-generated due to the user's English proficiency. This work is relevant to practitioners interested in model customization, layer insertion techniques, and open-source model modification.

Entities

TOTORONG(person)extGemma4-40_5B(model)Gemma 4(model)Google(company)Hugging Face(tool)

Related

CommunitySun, Jul 12, 2026, 04:30 AM

User seeks advice on adding second cheap GPU to run larger local models

A Reddit user running Gemma 4 26B-A4B at 12-15 t/s on an RTX 3060 (12 GB) finds the model insufficiently intelligent and wants to upgrade to a 31B model, which runs at only 1.5 t/s. They ask the community about the benefits of adding a second cheap GPU to improve performance for local LLM inference.

7 engagement·1 source·reddit
CommunitySat, Jul 11, 2026, 06:34 PM

Reddit user seeks fine-tuning wisdom from experienced practitioners

A Reddit user posted a request for practical fine-tuning advice from those who have fine-tuned more than half a model, seeking tips on dataset curation, LoRA rank selection, and cost debugging. The post emphasizes real-world experience over generic documentation.

3 engagement·1 source·reddit
Model ReleaseMon, Jul 6, 2026, 11:57 PM

Tencent Hy Team releases Hy3, a 295B-parameter MoE model

Tencent's Hy Team released Hy3, a 295-billion-parameter Mixture-of-Experts (MoE) language model with 21B active parameters and a 3.8B MTP layer. The model outperforms similar-sized models and rivals open-source models with 2-5x parameters. It is available on Hugging Face in full (598GB) and FP8 quantized (300GB) versions.

0 engagement·1 source·rss
RSS
CommunitySun, Jul 12, 2026, 05:03 AM

Developer compresses GLM-5.2 MoE to run on single RTX 3090 via 79 experiments

A developer conducted 79 experiments to compress GLM-5.2, a 337 GB MoE model with 75 sparse layers and 256 routed experts, to fit on a 24 GB RTX 3090. The approach uses per-expert codecs, a batch pipeline over all MoE layers, and a patched llama.cpp runtime that loads codec-native expert binaries at inference time. The MIT-licensed repository documents the method and findings on expert similarity.

1 engagement·1 source·reddit
ProductSun, Jul 12, 2026, 08:48 AM

Gemma 4 inference in Godot via GDScript and Vulkan compute shaders

An experimental Godot 4.7 project that runs Gemma 4 inference entirely in GDScript and Vulkan compute shaders. It enables LLM inference within the Godot game engine, potentially for in-game AI or interactive experiences.

1 engagement·1 source·hackernews