L3 Arcania 4x8b is an open-source language model by Steelskull. Features: 24.9b LLM, VRAM: 50GB, Context: 8K, License: llama3, MoE, Instruction-Based, HF Score: 67.5, LLM Explorer Score: 0.15, Arc: 62.1, HellaSwag: 82.7, MMLU: 66.9, TruthfulQA: 53.1, WinoGrande: 77.9, GSM8K: 62.6.
Autotrain compatible Conversational Endpoints compatible Instruct Mixtral Moe Not-for-all-audiences Region:us Safetensors Sharded Tensorflow
Model Card on HF 🤗: https://huggingface.co/Steelskull/L3-Arcania-4x8b
L3 Arcania 4x8b Benchmarks
ARC: 62.12 vs 96.7 (so35)-35.8%
HellaSwag: 82.72 vs 95.3 (gpt4)-13.2%
MMLU: 66.86 vs 88.3 (so35)-24.3%
TruthfulQA: 53.09 vs 59 (gpt4)-10%
WinoGrande: 77.9 vs 87.5 (gpt4)-11%
GSM8K: 62.55 vs 96.4 (so35)-35.1%
LLME Score: 0.14859
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
What is the LLM Explorer Rank (Score)
L3 Arcania 4x8b Parameters and Internals
| LLM Name | L3 Arcania 4x8b |
| Repository 🤗 | https://huggingface.co/Steelskull/L3-Arcania-4x8b |
| Model Size | 24.9b |
| Required VRAM | 50 GB |
| Updated | 2025-09-23 |
| Maintainer | Steelskull |
| Model Type | mixtral |
| Instruction-Based | Yes |
| Model Files | |
| Model Architecture | MixtralForCausalLM |
| License | llama3 |
| Context Length | 8192 |
| Model Max Length | 8192 |
| Transformers Version | 4.40.1 |
| Tokenizer Class | PreTrainedTokenizerFast |
| Padding Token | <|begin_of_text|> |
| Vocabulary Size | 128256 |
| Torch Data Type | bfloat16 |
Best Alternatives to L3 Arcania 4x8b
Best Alternatives |
Context / RAM |
Downloads |
Likes |
|---|---|---|---|
| L3.1 MoE 4x8B V0.2 | 128K / 50.1 GB | 7 | 2 |
| OpenCrystal V4 L3 4x8B | 8K / 50 GB | 5 | 2 |
| Llama 3 8B Instruct MoE 2 | 8K / 50 GB | 10 | 0 |
| Llama 3 4x8B | 8K / 49.9 GB | 7 | 0 |
Note: green Score (e.g. "73.2") means that the model is better than Steelskull/L3-Arcania-4x8b .
Rank the L3 Arcania 4x8b Capabilities
🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟
What open-source LLMs or SLMs are you in search of? 52721 in total.
Original data from HuggingFace, OpenCompass and various public git repos.
Check out Ag3ntum — our secure, self-hosted AI agent for server management.
Release v20260328a