Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google releases Gemma 4 open models

deepmind.google

Post reply on HN
Print view View original

351–360 of 507 posts

Re: Google releases Gemma 4 open models

#351

Post by mh- »

mentalgear wrote: post #82
adamtaylor_13 wrote: post #22

Earlier quoted context omitted.

What sort of tasks are you using self-hosting for? Just curious as I've been watching the scene but not experimenting with self-hosting.

Adding to the Q: Any good small open-source model with a high correctness of reading/extracting Tables and/of PDFs with more uncommon layouts.

I haven't tried it yet, but I bookmarked this recently: https://github.com/opendataloader-project/opendataloader-pdf

Re: Google releases Gemma 4 open models

#353

Post by coder543 »

simonw wrote: post #342

Anyone figured out a recipe to run Gemma 4 E2B or E4B against audio files locally on a Mac?

If you search the model card[0], there is a section titled "Code for processing Audio", which you can probably use to test things out. But, the model card makes the audio support seem disappointing:

> Audio supports a maximum length of 30 seconds.

[0]: https://huggingface.co/google/gemma-4-26B-A4B-it#getting-sta...

Re: Google releases Gemma 4 open models

#354

Post by c0wb0yc0d3r »

yags wrote: post #348
simonw wrote: post #77

I ran these in LM Studio and got unrecognizable pelicans out of the 2B and 4B models and an outstanding pelican out of the 26b-a4b model - I think the best I've seen from a model that runs on my laptop. https://simonwillison.net/2026/Apr/2/gemma-4/ The gemma-4-31b model is completely broken for me - it just spits out "---\n" no matter what prompt I feed it. I got a pelican out of it via the AI Studio API hosted model…

We (LM Studio) found the bug with the 31B model and a fix will be going out hopefully tonight

I am not deep in this world. What does it mean when you (LM Studio) fixed a bug in a model Google released?

Re: Google releases Gemma 4 open models

#355

Post by rahimnathwani »

simonw wrote: post #342

Anyone figured out a recipe to run Gemma 4 E2B or E4B against audio files locally on a Mac?

Prince Canuma just updated mlx-vlm: https://x.com/i/status/2039815307821199709

So something like this should work: https://x.com/i/status/1938328542699503723

Re: Google releases Gemma 4 open models

#356

Post by daemonologist »

swalsh wrote: post #341
redman25 wrote: post #280

Earlier quoted context omitted.

Exactly, compare MoE with MoE and dense with dense otherwise it's apples and oranges.

Its coding to coding. I could care less how the model is architected, i only care how it performs in a real world scenario.

The implication is that there is (should be) a major speed difference - naively you'd expect the MoE to be 10x faster and cheaper, which can be pretty relevant on real world tasks.

Re: Google releases Gemma 4 open models

#358

Post by mh- »

jeffbee wrote: post #208
hrmtst93837 wrote: post #198

Earlier quoted context omitted.

[flagged]

In my experience the contents of the message are all but totally irrelevant to the classification, and it is the behavior of the mailing peer that gives all the relevant features.

Based on how much blatant gmail->gmail spam I receive, the gmail team agrees with this strategy.

Re: Google releases Gemma 4 open models

#359

Post by Retro_Dev »

I'm very pleased with the performance of the largest gemma4 model (which I tested through ollama). My singular data point on whether an LLM remembers things well is whether it can translate toki pona to (and from) English. I find it easy to evaluate because I know the language. This local LLM marks the first version that 1) doesn't hallucinate words - at least, for the largest model - and 2) uses common word-phrases that other toki pona speakers use, and most importantly 3) can actually run on my laptop.
Post reply on HN