Skip to content
NEWLive updates feed from Hugging Face, arXiv and GitHub  See what changed →

farbodtavakkoli · gemma4 · based on google/gemma-4-31b-it

OTel-2.0-LLM-31B-IT

OTel-2.0-LLM-31B-IT by farbodtavakkoli: a 31.3B-parameter dense open-weight model under the apache-2.0 license and vision.

apache-2.0Dense262K contextVisionAuto-imported · not yet reviewed

6,916,354 downloads · 20 likes · Model card · synced 2026-09-19

Specifications

Parameters
31.3B
Architecture
dense · gemma4
Layers
60
Hidden size
5,376
KV heads · head dim
16 · 256
Vocabulary
262,144
Native dtype
Context window
262,144 tokens
Released
2026-07-23
Last modified on Hub
2026-09-08

Features & licensing

License
apache-2.0
Commercial use
Yes
Modalities
text, vision
Tool / function calling
No
Reasoning mode
No
Languages
EN
Quantised variants on Hub
FP8, GGUF
Pipeline
text-generation
transformerssafetensorsgemma4image-text-to-texttelecomtelecommunicationsgsmaat-tmicrosoftdellamdred-hatopen-telco-airaginstruction-followingtool-callingosftorthogonal-subspace-fine-tuningcontinual-learningdomain-adaptationpost-trainedvllmmultimodaltext-generationconversationalenendpoints_compatible

Task fit

Editorial scores (0–100) used by the recommendation engine.

  • Chat70
  • RAG / Q&A70
  • Code70
  • Summarisation70
  • Extraction70
  • Agents70

Hardware to run OTel-2.0-LLM-31B-IT

Weights need about 69 GB at FP16, 36 GB at INT8 and 21 GB at INT4. The KV cache adds roughly 983 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 58.8 GB, which fits on 1 × NVIDIA A100 (80 GB).

GPU configurations at the reference workload
ConfigurationVRAMUtilisationEst. first tokenCloud / month
3 × NVIDIA L4 (24 GB)72 GB82 %~0.5 s$1,752
2 × NVIDIA L40S (48 GB)96 GB61 %~0.2 s$2,774
1 × NVIDIA A100 (80 GB)recommended80 GB73 %~0.1 s$2,336
1 × NVIDIA H100 (80 GB)80 GB73 %~0.0 s$3,285

Estimates only. Run an assessment for your own traffic.

Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.