google/gemma-4-12B-it · Discussions

#50 opened 12 days ago by

atbolsh

fix: chat_template, keep tool-call reasoning across later user turns

👍 1

#49 opened 14 days ago by

dimondev

Raise per-image vision soft-token budget from 280 to 1120

#46 opened about 1 month ago by

lucianommartins

Gemma 4 12B Unified (encoder-free audio): audio comprehension collapses past ~450–500 tokens of absolute position — vision on the same checkpoint is unaffected, E2B/E4B (Conformer) are unaffected

#45 opened about 1 month ago by

frandagostino

SAE for this model?

#44 opened about 1 month ago by

joggyback

https://share.gemini.google/7T1rHMOMMfPH

#42 opened about 1 month ago by

DIGITALLEADERSHIP

gemma-4-12B-it: deterministic "thought\n thought\n …" degenerate loop on long agent prompts (~60% reproducer, 4-bit, repros across temperatures)

9

#41 opened about 2 months ago by

Raullen

Fairly Useless

👍 3

#40 opened about 2 months ago by

yano2mch

Gemma 4 12b bias affraid me

#39 opened 2 months ago by

Alekto

Chat template may re-inject prior-turn reasoning during multi-turn tool use → repetition loops

8

#38 opened 2 months ago by

ManniX-ITA

fdsfsdfsdfdf

#37 opened 2 months ago by

jiwang1

asfd

#36 opened 2 months ago by

jiwang1

Gemma 4 124b, please!

🚀🔥 9

#34 opened 2 months ago by

slavczays

vllm support?

#32 opened 2 months ago by

mohamedemam

Gemma4-12B Audio Deep Evaluation

#30 opened 2 months ago by

PeterchanCN

How to use an assistent model for MTP in llama.cpp?

#29 opened 2 months ago by

Regrin

Tool Calling and Tags and Context awareness.

👍 1

#28 opened 2 months ago by

JanjanJean

MLX 4-bit conversion of gemma-4-12B-it — with vision + audio embedder weights preserved

#27 opened 2 months ago by

jokernifty

Gemma 4 124B DENSE + Nvidia N1x Laptop

❤️🔥 2

#26 opened 2 months ago by

87455a

What languages are supported in audio input?

🤝 1

#25 opened 2 months ago by

J22

it is speaking Cantonese while user prompt is Simplified Chinese

#24 opened 2 months ago by

leiyang88

Can't deploy on huggingface...

#23 opened 2 months ago by

p-christ

NVFP4 Quant

#22 opened 2 months ago by

berkerdooo

👋 Training support for transformers/megatron backends

#20 opened 2 months ago by

study-hjt

NVFP4 / FP8 Quantizations

#19 opened 2 months ago by

vincentzed-hf

Dimension mismatch crashes the engine at startup

👍 2

#18 opened 2 months ago by

CallShaul

Gemma 4 124b

🚀 20

10

#17 opened 2 months ago by

FusionCow

Thank you Google team

❤️🔥 5

#16 opened 2 months ago by

jreoka

Thanks again!

❤️ 1

#15 opened 2 months ago by

JanjanJean

MTP Assistant Model

#14 opened 2 months ago by

patrickdeanbrown

Preserve the correct ordering of content and tool calls when rendering OpenAI Chat Completions

👀 2

#12 opened 2 months ago by

yzong-rh

Good, Google

🤝👍 4

5

#9 opened 2 months ago by

dof-studio-org

Add HLE evaluation result

#7 opened 2 months ago by

Add AIME 2026 evaluation result

#6 opened 2 months ago by

Add MMMU Pro evaluation result

#5 opened 2 months ago by

Add MMLU-Pro evaluation result

#4 opened 2 months ago by

Add GPQA Diamond evaluation result

#3 opened 2 months ago by

This is an awesome model proposition/idea for users with few VRAM.

🚀 2