Moonlight 16B A3B SFT EchoMoE Teacher is an open-source language model by McGill-NLP. Features: 16b LLM, VRAM: 32.6GB, Context: 8K, License: mit, Instruction-Based.
| LLM Name | Moonlight 16B A3B SFT EchoMoE Teacher |
| Repository π€ | https://huggingface.co/McGill-NLP/Moonlight-16B-A3B-SFT-EchoMoE-teacher |
| Base Model(s) | |
| Model Size | 16b |
| Required VRAM | 32.6 GB |
| Updated | 2026-11-19 |
| Maintainer | McGill-NLP |
| Model Type | deepseek_v3 |
| Instruction-Based | Yes |
| Model Files | |
| Supported Languages | en |
| Model Architecture | DeepseekV3ForCausalLM |
| License | mit |
| Context Length | 8192 |
| Model Max Length | 8192 |
| Transformers Version | 5.8.1 |
| Tokenizer Class | PreTrainedTokenizerFast |
| Padding Token | [PAD] |
| Vocabulary Size | 163840 |
Best Alternatives |
Context / RAM |
Downloads |
Likes |
|---|---|---|---|
| Moon Beam 16B A3B | 8K / 32.6 GB | 935 | 0 |
| Moonlight 16B A3B Instruct | 8K / 32.6 GB | 45597 | 196 |
| ...3B Instruct Abliterated Pruned | 8K / 17.5 GB | 81 | 6 |
| ...onlight 16B A3B Instruct 4 Bit | 8K / 9 GB | 136 | 4 |
| ...t 16B A3B Instruct Abliterated | 8K / 31.9 GB | 16 | 8 |