LLM EXPLORER 58,297 MODELS INDEXED

Jamba 900M V0.13 KIx2 by pszemraj

By pszemraj · 20 downloads
HF Repo ↗ Copy URL

Jamba 900M V0.13 KIx2 is an open-source language model by pszemraj. Features: 900m LLM, VRAM: 1.8GB, Context: 16K, License: apache-2.0, LLM Explorer Score: 0.12.

Dataset:bee-spoke-data/fineweb... Dataset:bee-spoke-data/fineweb... Dataset:bee-spoke-data/fineweb... Dataset:bee-spoke-data/knowled... Dataset:bee-spoke-data/wikiped...   En   Jamba   Region:us   Safetensors   Smol   Smol moe
Model Card on HF 🤗: https://huggingface.co/pszemraj/jamba-900M-v0.13-KIx2 

Jamba 900M V0.13 KIx2 Parameters and Internals

Model Type 
multimodal, text generation
Additional Notes 
The API widget is off as it isn't supported by hf yet - try the Colab.
Supported Languages 
en (primary)
Training Details 
Data Sources:
BEE-spoke-data/knowledge-inoc-concat-v1, BEE-spoke-data/wikipedia-20230901.en-deduped, BEE-spoke-data/fineweb-100k_en-med, BEE-spoke-data/fineweb-1M_en-med, BEE-spoke-data/fineweb-1M_longish
Data Volume:
20b tokens
Methodology:
pretraining as smol MoE with Claude3 tokenizer
Context Length:
16384
Model Architecture:
hidden size 1024, 12 layers, 8 experts
Input Output 
Input Format:
Claude3 tokenizer format
Accepted Modalities:
text
Output Format:
text
LLM NameJamba 900M V0.13 KIx2
Repository 🤗https://huggingface.co/pszemraj/jamba-900M-v0.13-KIx2 
Model Size900m
Required VRAM1.8 GB
Updated2026-08-08
Maintainerpszemraj
Model Typejamba
Model Files  1.8 GB   0.0 GB
Supported Languagesen
Model ArchitectureJambaForCausalLM
Licenseapache-2.0
Context Length16384
Model Max Length16384
Transformers Version4.40.1
Tokenizer ClassGPT2Tokenizer
Vocabulary Size65024
Torch Data Typebfloat16

Best Alternatives to Jamba 900M V0.13 KIx2

Best Alternatives
Context / RAM
Downloads
Likes
Jamba 900M Doc Writer16K / 1.8 GB83
Note: green Score (e.g. "73.2") means that the model is better than pszemraj/jamba-900M-v0.13-KIx2.