LLM EXPLORER 58,297 MODELS INDEXED

Mega Ar 525M V0.07 UltraTBfw by pszemraj

By pszemraj · 7 downloads
HF Repo ↗ Copy URL

Mega Ar 525M V0.07 UltraTBfw is an open-source language model by pszemraj. Features: 525m LLM, VRAM: 2.1GB, License: apache-2.0, LLM Explorer Score: 0.12.

Dataset:bee-spoke-data/ultrate...   En   Endpoints compatible   Mega   Region:us   Safetensors
Model Card on HF 🤗: https://huggingface.co/pszemraj/mega-ar-525m-v0.07-ultraTBfw 

Mega Ar 525M V0.07 UltraTBfw Parameters and Internals

Model Type 
text generation
Additional Notes 
The model is fine-tuned for enhanced performance on the specified dataset, targeting diverse language tasks as evidenced by the evaluation benchmarks.
Supported Languages 
en (full support)
Training Details 
Data Sources:
BEE-spoke-data/UltraTextbooks-2.1-fw_mix
Methodology:
fine-tuning
Context Length:
4096
Model Architecture:
1536 hidden size, 3x hidden:FF, 8 layers, MEGA EMA dim 32, llama-3 tokenizer
LLM NameMega Ar 525M V0.07 UltraTBfw
Repository 🤗https://huggingface.co/pszemraj/mega-ar-525m-v0.07-ultraTBfw 
Model Size525m
Required VRAM2.1 GB
Updated2026-07-30
Maintainerpszemraj
Model Typemega
Model Files  2.1 GB   0.0 GB
Supported Languagesen
Model ArchitectureMegaForCausalLM
Licenseapache-2.0
Transformers Version4.40.2
Tokenizer ClassPreTrainedTokenizerFast
Vocabulary Size128256
Torch Data Typefloat32