Model overview

gemma-4-31b-it appears in the Google section of NVIDIA's model catalogue as a dense 31B model. Its stated focus is frontier reasoning applied to coding, agentic workflows and fine-tuning. Readers searching for the gemma 4 31b instruct model will find it listed alongside other Google entries in the same catalogue.

  • Dense 31B parameter model
  • Publisher listed as Google
  • Positioned for coding and agentic workflows
  • Supports fine-tuning

Source: AI Models by Google | Try NVIDIA NIM APIs

Capability tags

The catalogue entry carries tags that indicate the model's intended modality and behaviour. These are text-to-text, reasoning and agentic, placing it among language models rather than vision or diffusion models.

  • text-to-text
  • reasoning
  • agentic

Source: AI Models by Google | Try NVIDIA NIM APIs

Availability and access

The listing marks gemma-4-31b-it as downloadable and offered with a free endpoint, so it can be tried through the NVIDIA NIM APIs or obtained for local use. Access is provided from the Google model page on NVIDIA's build site.

  • Downloadable
  • Free endpoint
  • Accessed via NVIDIA NIM APIs

Source: AI Models by Google | Try NVIDIA NIM APIs

Related Google models in the catalogue

The Google page lists three models in total. Besides gemma-4-31b-it, it includes diffusiongemma-26b-a4b-it, a diffusion-based 26B parameter LLM for parallel token generation in real-time text applications, and paligemma, a vision language model that handles both text and visual inputs.

  • diffusiongemma-26b-a4b-it — diffusion-based 26B LLM
  • paligemma — vision language model

Source: AI Models by Google | Try NVIDIA NIM APIs

Sources

  1. AI Models by Google | Try NVIDIA NIM APIs https://build.nvidia.com/google Verified 03 Nov 2026

Last verified 02 Oct 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.