LLM EXPLORER 59,358 MODELS INDEXED

Granite 3.0 8B Base by ibm-granite

By ibm-granite · 5202 downloads
HF Repo ↗ Copy URL

Granite 3.0 8B Base is an open-source language model by ibm-granite. Features: 8b LLM, VRAM: 32.8GB, Context: 4K, License: apache-2.0, LLM Explorer Score: 0.24.

  Arxiv:0000.00000   Eval-results   Granite   Granite-3.0   Language   Model-index   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/ibm-granite/granite-3.0-8b-base 

Granite 3.0 8B Base Parameters and Internals

Model Type 
text-generation
Use Cases 
Areas:
text-to-text generation
Applications:
summarization, text classification, extraction, question-answering
Primary Use Cases:
baseline to create specialized models
Limitations:
Not undergone any safety alignment, may produce problematic outputs., Potential increased susceptibility to hallucination due to model size.
Considerations:
Community urged to use the model with ethical intentions.
Supported Languages 
English (supported), German (supported), Spanish (supported), French (supported), Japanese (supported), Portuguese (supported), Arabic (supported), Czech (supported), Italian (supported), Korean (supported), Dutch (supported), Chinese (supported)
Training Details 
Data Sources:
web, code, academic sources, books, math data, multilingual, instruction data
Data Volume:
12 trillion tokens for Stage 1 and 2 trillion tokens for Stage 2
Methodology:
Two-stage training strategy
Context Length:
4096
Hardware Used:
IBM's super computing cluster, Blue Vela, NVIDIA H100 GPUs
Model Architecture:
Decoder-only dense transformer architecture with GQA, RoPE, MLP with SwiGLU, RMSNorm, and shared input/output embeddings
Responsible Ai Considerations 
Fairness:
Involves awareness of bias and fairness.
Mitigation Strategies:
Ongoing research to address and mitigate issues.
Input Output 
Input Format:
Tokenized input using AutoTokenizer
Accepted Modalities:
text
Output Format:
Decodes output tokens into text using AutoTokenizer
Performance Tips:
Use appropriate libraries and follow examples provided.
LLM NameGranite 3.0 8B Base
Repository 🤗https://huggingface.co/ibm-granite/granite-3.0-8b-base 
Model Size8b
Required VRAM32.8 GB
Updated2026-08-01
Maintaineribm-granite
Model Typegranite
Model Files  4.8 GB: 1-of-7   4.8 GB: 2-of-7   4.8 GB: 3-of-7   4.8 GB: 4-of-7   4.8 GB: 5-of-7   4.8 GB: 6-of-7   4.0 GB: 7-of-7
Model ArchitectureGraniteForCausalLM
Licenseapache-2.0
Context Length4096
Model Max Length4096
Transformers Version4.45.0.dev0
Tokenizer ClassGPT2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size49152

Best Alternatives to Granite 3.0 8B Base

Best Alternatives
Context / RAM
Downloads
Likes
Granite 4.1 8B128K / 17.6 GB21803161
Granite 3.1 8B Instruct128K / 16.4 GB138888168
Granite 4.1 8B Base128K / 16.8 GB1134825
Granite Guardian 4.1 8B128K / 16.8 GB3563733
Granite 3.1 8B Reasoning128K / 16.4 GB51
Granite 3.1 8B Instruct128K / 16.4 GB1851
... Granite 4.1 8B Claude Fable 5128K / 33.6 GB8465
Granite 3.2 8B Instruct128K / 16.4 GB379292
Granite 4.1 8B SFT Fable5128K / 17.6 GB2370
...NG Coder V11.2800 Heretic BF16128K / 16.8 GB1411
Note: green Score (e.g. "73.2") means that the model is better than ibm-granite/granite-3.0-8b-base.