91–100 of 507 posts
Earlier quoted context omitted.
Will larger-parameter versions be released?
We are always figuring out what parameter size makes sense. The decision is always a mix between how good we can make the models from a technical aspect, with how good they need to be to make all of you super excited to use them. And its a bit of a challenge what is an ever changing ecosystem. I'm personally curious is there a certain parameter size you're looking for?
Was it too good or not good enough? (blink twice if you can't answer lol)
Featuring the ELO score as the main benchmark in chart is very misleading. The big dense Gemma 4 model does not seem to reach Qwen 3.5 27B dense model in most benchmarks. This is obviously what matters. The small 2B / 4B models are interesting and may potentially be better ASR models than specialized ones (not just for performances but since they are going to be easily served via llama.cpp / MLX and front-ends). Also…
Public benchmarks can be trivially faked. Lmarena is a bit harder to fake and is human-evaluated. I agree it's misleading for them to hyper-focus on one metric, but public benchmarks are far from the only thing that matters. I place more weight on Lmarena scores and private benchmarks.
Thinking / reasoning + multimodal + tool calling. We made some quants at https://huggingface.co/collections/unsloth/gemma-4 for folks to run them - they work really well! Guide for those interested: https://unsloth.ai/docs/models/gemma-4 Also note to use temperature = 1.0, top_p = 0.95, top_k = 64 and the EOS is " ". " thought\n" is also used for the thinking trace!
You have an answer on your page regarding "Should I pick 26B-A4B or 31B?", but can you please clarify if, assuming 24GB vRAM, I should pick a full precision smaller model or 4 bit larger model?
Hi all! I work on the Gemma team, one of many as this one was a bigger effort given it was a mainline release. Happy to answer whatever questions I can
How is the performance for Japanese, voice in particular?
The benchmark comparisons to Gemma 3 27B on Hugging Face are interesting: The Gemma 4 E4B variant ( https://huggingface.co/google/gemma-4-E4B-it ) beats the old 27B in every benchmark at a fraction of parameters. The E2B/E4B models also support voice input, which is rare.
Thinking vs non-thinking. There'll be a token cost there. But still fairly remarkable!
If you want the fastest open source implementation on Blackwell and AMD MI355, check out Modular's MAX nightly. You can pip install it super fast, check it out here: https://www.modular.com/blog/day-zero-launch-fastest-perform... -Chris Lattner (yes, affiliated with Modular :-)
Earlier quoted context omitted.
Not sure why you're being downvoted, the other thing Google has is Google. They just have to spend the effort/resources to keep up and wait for everyone else to go bankrupt. At the end of the day I think Google will be the eventual LLM winner. I think this is why Meta isn't really in the race and just releases open weight models, the writing is on the wall. Also, probably why Apple went ahead and signed a deal with G…
I don't know why I am downvoted but Google has data, expertise, hardware and deep pockets. This whole LLM thing is invented at Google and machine learning ecosystem libraries come from Google. I don't know how people can be so irrational discounting Google's muscle. Others have just borrowed data, money, hardware and they would run out of resources for sure.
Earlier quoted context omitted.
We are always figuring out what parameter size makes sense. The decision is always a mix between how good we can make the models from a technical aspect, with how good they need to be to make all of you super excited to use them. And its a bit of a challenge what is an ever changing ecosystem. I'm personally curious is there a certain parameter size you're looking for?
Mainline consumer cards are 16GB, so everyone wants models they can run on their $400 GPU.