Google announces open Gemma 4 model with Apache 2.0 license - 9to5Google
Google announces open Gemma 4 model with Apache 2.0 license 9to5Google
Could not retrieve the full article text.
Read on GNews AI Gemma →Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
modelannounce
How well do current models handle Icelandic audio?
I’ve been doing some informal testing on how current multimodal models handle speech + multilingual understanding, and came across an interesting behavior that feels slightly beyond standard translation.I used a short audio clip in a language I don’t understand (likely Icelandic) and evaluated the output along a few dimensions:1. Transcription qualityThe model produced a relatively clean transcript, with no obvious structural breakdown.2. Translation fidelity vs. fluencyInstead of sticking closely to literal phrasing, the translation leaned more toward natural English, sometimes smoothing or rephrasing content.3. Context / tone inferenceThis was the most notable part — the model attempted to describe the tone and intent of the speakers (e.g., casual vs. serious), which goes beyond typical
Spent the weekend reading a local agent runtime repo. The TS-only packaging and persistent MCP ports are both very smart.
I like reading local LLM infra repos more than launch posts, and I ended up deep in one this weekend because it supports local providers like Ollama. Two things gave me the “okay, someone actually cared about runtime engineering” reaction. First, the runtime path was moved fully into TypeScript. The API layer, runner orchestration, workspace MCP hosting, and packaging all live there now, and the packaged runtime no longer ships Python source or Python deps. For local/self-hosted stacks that matters more than it sounds: smaller bundle, fewer moving pieces, less cross-language drift. Second, they stopped doing hardcoded MCP port math. Ports are persisted in SQLite with UNIQUE(port) and (workspace_id, app_id) as the key, and the runner merges prepared MCP servers during bootstrap. So local si
Per-Layer Embeddings: A simple explanation of the magic behind the small Gemma 4 models
Many of you seem to have liked my recent post "A simple explanation of the key idea behind TurboQuant" . Now I'm really not much of a blogger and I usually like to invest all my available time into developing Heretic, but there is another really cool new development happening with lots of confusion around it, so I decided to make another quick explainer post. You may have noticed that the brand-new Gemma 4 model family includes two small models: gemma-4-E2B and gemma-4-E4B . Yup, that's an "E", not an "A". Those are neither Mixture-of-Experts (MoE) models, nor dense models in the traditional sense. They are something else entirely, something that enables interesting new performance tradeoffs for inference. What's going on? To understand how these models work, and why they are so cool, let'
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!