LLM EXPLORER 59,420 MODELS INDEXED

Minotaur 15B by openaccess-ai-collective

By openaccess-ai-collective · 69 downloads
HF Repo ↗ Copy URL

Minotaur 15B is an open-source language model by openaccess-ai-collective. Features: 15b LLM, VRAM: 31.2GB, Instruction-Based, Code Generating, LLM Explorer Score: 0.07.

  Arxiv:1911.02150   Arxiv:2205.14135   Arxiv:2207.14255   Arxiv:2305.06161   Code   Codegen Dataset:bigcode/the-stack-dedu...   Dataset:camel-ai/biology   Dataset:camel-ai/chemistry   Dataset:camel-ai/math   Dataset:camel-ai/physics Dataset:ehartford/wizardlm alp...   Dataset:gsm8k   Dataset:hellaswag Dataset:metaeval/scienceqa tex... Dataset:openai/summarize from ...   Dataset:qingyisi/alpaca-cot   Dataset:riddle sense Dataset:teknium/gpteacher-gene... Dataset:tiiuae/falcon-refinedw...   Dataset:winglian/evals   Endpoints compatible   Gpt bigcode   Instruct   Pytorch   Region:us   Sharded
Model Card on HF 🤗: https://huggingface.co/openaccess-ai-collective/minotaur-15b 

Minotaur 15B Parameters and Internals

Model Type 
text generation
Use Cases 
Areas:
research, commercial applications
Applications:
prose generation, programming assistance
Primary Use Cases:
programming tasks, academic research
Limitations:
not aligned to human preferences, may produce problematic outputs, limitations with non-English text
Considerations:
Use generated code with caution
Additional Notes 
The model is instruct fine-tuned and can produce code verbatim from the dataset.
Supported Languages 
English (proficient), programming_languages (>80)
Training Details 
Data Sources:
bigcode/the-stack-dedup, tiiuae/falcon-refinedweb, ehartford/WizardLM_alpaca_evol_instruct_70k_unfiltered, QingyiSi/Alpaca-CoT, teknium/GPTeacher-General-Instruct, metaeval/ScienceQA_text_only, openai/summarize_from_feedback, riddle_sense, gsm8k, camel-ai/math, camel-ai/biology, camel-ai/physics, camel-ai/chemistry, winglian/evals
Data Volume:
600B fine-tuning tokens
Methodology:
fine-tuning with instruction data
Context Length:
8192
Training Time:
approximately 30 hours on 4xA100 80GB for 1 epoch
Hardware Used:
4xA100 80GB GPUs
Model Architecture:
GPT-2 model with multi-query attention and Fill-in-the-Middle objective
LLM NameMinotaur 15B
Repository 🤗https://huggingface.co/openaccess-ai-collective/minotaur-15b 
Model Size15b
Required VRAM31.2 GB
Updated2026-07-27
Maintaineropenaccess-ai-collective
Model Typegpt_bigcode
Instruction-BasedYes
Model Files  10.0 GB: 1-of-4   9.9 GB: 2-of-4   9.9 GB: 3-of-4   1.4 GB: 4-of-4
Generates CodeYes
Model ArchitectureGPTBigCodeForCausalLM
Transformers Version4.28.1
Tokenizer ClassGPT2Tokenizer
Vocabulary Size49152
Torch Data Typefloat32
Activation Functiongelu

Best Alternatives to Minotaur 15B

Best Alternatives
Context / RAM
Downloads
Likes
WizardCoderSQL 15B V1.00K / 31.2 GB61
...r Skeleton Wizard Coder Merged0K / 31.2 GB1813
...der Natsql Wizard Coder Merged0K / 32.7 GB21
Minotaur 15B GPTQ0K / 9.2 GB68
Note: green Score (e.g. "73.2") means that the model is better than openaccess-ai-collective/minotaur-15b.