Skills2026-04-11 · source-backed

Re-download Gemma 4 GGUFs if you got them before April 10.

Issue

2026-04-11

Confidence

source-backed

Sources

Redaction

passed

Story

Unsloth updated all uploads with corrected chat templates, explicit <|think|> thinking control, and fixes for grad accumulation exploding losses (300-400 down to 10-15). Old versions have broken chat behavior and a 26B/31B inference IndexError.

Ollama supports Gemma / Shared entities / Same source domain / Shared topic / Earlier coverage
Gemma 4 Drops Four Open Models, and the 31B Dense Beats a 397B MoE
Linked by a graph relationship (Ollama supports Gemma); both cover April, Gemma, LocalLLaMA, Unsloth; reported by the same outlet (reddit.com).
Ollama supports Gemma / Shared entities / Same source domain / Earlier coverage
Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.
Linked by a graph relationship (Ollama supports Gemma); both cover April, Gemma, LocalLLaMA; reported by the same outlet (reddit.com).
Ollama supports Gemma / Shared entities / Shared topic / What happened next
Gemma 4 QAT checkpoints drop the smallest model from 11.4GB to 1.1GB
Linked by a graph relationship (Ollama supports Gemma); both cover Gemma, Unsloth; overlapping topics (gemma, gguf, inference).
Unsloth supports DeepSeek / Shared entities / What happened next
Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.
Linked by a graph relationship (Unsloth supports DeepSeek); both cover April, Gemma, LocalLLaMA; picks up the April thread on 2026-05-01.
Unsloth supports DeepSeek / Shared entities / Same source domain / What happened next
Meta ships Muse Glimmer, a 30B open agent model
Linked by a graph relationship (Unsloth supports DeepSeek); both cover April, LocalLLaMA; reported by the same outlet (reddit.com).
Ollama supports Gemma / Shared entities / Same source domain / What happened next / Tension
A 27B Model Just Matched the 400B One. It Fits on Your Laptop.
Linked by a graph relationship (Ollama supports Gemma); both cover April, LocalLLaMA, Unsloth; reported by the same outlet (reddit.com).
Unsloth supports Blackwell / Shared entity: LocalLLaMA / Same source domain / Earlier coverage / Downstream implication
llama.cpp Gets Backend-Agnostic Tensor Parallelism, and the NVIDIA Tax Just Got Optional
Linked by a graph relationship (Unsloth supports Blackwell); both cover LocalLLaMA; reported by the same outlet (reddit.com).
Unsloth supports DeepSeek / Shared entities / Earlier coverage
Nvidia's Blackwell Ultra Just Made Self-Hosted Inference a Real Option: 2.49M Tokens/Sec, 30 Cents Per Million
Linked by a graph relationship (Unsloth supports DeepSeek); both cover April, Gemma; earlier April coverage from 2026-04-03.

Source trail

Unsloth updated all uploads

Entities

April Gemma GGUFs IndexError LocalLLaMA Unsloth

Provenance

Canonical issue: Ramsay Research Agent — April 11, 2026
AI generated: no
Story unit: 2026-04-11-re-download-gemma-4-ggufs-if-you-got-them-before-april-10
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Re-download Gemma 4 GGUFs if you got them before April 10.

Story

Related stories

Gemma 4 Drops Four Open Models, and the 31B Dense Beats a 397B MoE

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Gemma 4 QAT checkpoints drop the smallest model from 11.4GB to 1.1GB

Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.

Meta ships Muse Glimmer, a 30B open agent model

A 27B Model Just Matched the 400B One. It Fits on Your Laptop.

llama.cpp Gets Backend-Agnostic Tensor Parallelism, and the NVIDIA Tax Just Got Optional

Nvidia's Blackwell Ultra Just Made Self-Hosted Inference a Real Option: 2.49M Tokens/Sec, 30 Cents Per Million

Source trail

Entities

Provenance

Re-download Gemma 4 GGUFs if you got them before April 10.

Story

Related stories

Gemma 4 Drops Four Open Models, and the 31B Dense Beats a 397B MoE

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Gemma 4 QAT checkpoints drop the smallest model from 11.4GB to 1.1GB

Xiaomi Just Open-Sourced a 1-Trillion Parameter Model Under MIT License. It Built a Compiler in 4 Hours.

Meta ships Muse Glimmer, a 30B open agent model

A 27B Model Just Matched the 400B One. It Fits on Your Laptop.

llama.cpp Gets Backend-Agnostic Tensor Parallelism, and the NVIDIA Tax Just Got Optional

Nvidia's Blackwell Ultra Just Made Self-Hosted Inference a Real Option: 2.49M Tokens/Sec, 30 Cents Per Million

Source trail

Entities

Provenance