LLM EXPLORER 62,751 MODELS INDEXED

Swift1.5 Qwen3.8 Flash Next W4A16 Merlin by halt95

By halt95 · 64 downloads
HF Repo ↗ Copy URL

Swift1.5 Qwen3.8 Flash Next W4A16 Merlin is an open-source language model by halt95. Features: 124b LLM, VRAM: 71.4GB, License: other, LLM Explorer Score: 0.3.

Base model:halt95/qwen3.8-flas... Base model:merge:halt95/qwen3.... Base model:merge:ukisai/swift1... Base model:ukisai/swift1.5-qwe...   Compressed-tensors   Conversational   Flash-next   Fp8   Image-text-to-text   Int4   Int8   Mixed-precision   Mtp   Qwen3.8   Qwen4 exp   Region:us   Rtx-3090   Safetensors   Sharded   Tensorflow   Vllm
Model Card on HF 🤗: https://huggingface.co/halt95/Swift1.5-Qwen3.8-Flash-Next-W4A16-Merlin 

Swift1.5 Qwen3.8 Flash Next W4A16 Merlin Parameters and Internals

LLM NameSwift1.5 Qwen3.8 Flash Next W4A16 Merlin
Repository 🤗https://huggingface.co/halt95/Swift1.5-Qwen3.8-Flash-Next-W4A16-Merlin 
Base Model(s)  ukisai/Swift1.5-Qwen3.8-Flash-Next   halt95/Qwen3.8-Flash-Next-W4A16-Merlin   ukisai/Swift1.5-Qwen3.8-Flash-Next   halt95/Qwen3.8-Flash-Next-W4A16-Merlin
Model Size124b
Required VRAM71.4 GB
Updated2026-10-02
Maintainerhalt95
Model Typeqwen4_exp
Model Files  2.8 GB: 1-of-17   5.2 GB: 3-of-17   5.2 GB: 4-of-17   5.2 GB: 5-of-17   5.2 GB: 6-of-17   5.2 GB: 7-of-17   5.2 GB: 8-of-17   5.2 GB: 9-of-17   5.2 GB: 10-of-17   5.1 GB: 11-of-17   5.2 GB: 12-of-17   5.2 GB: 13-of-17   5.2 GB: 14-of-17   5.0 GB: 15-of-17   1.3 GB: 17-of-17   5.2 GB   5.2 GB   5.2 GB   5.2 GB   5.2 GB   5.2 GB   5.2 GB   5.2 GB   5.2 GB   4.4 GB   0.2 GB   1.3 GB
Model ArchitectureQwen4ExpForConditionalGeneration
Licenseother
Model Max Length262144
Transformers Version5.8.0.dev0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Swift1.5 Qwen3.8 Flash Next W4A16 Merlin

Best Alternatives
Context / RAM
Downloads
Likes
...wen3.8 Flash Next W4A16 Merlin0K / 71.4 GB13138
Note: green Score (e.g. "73.2") means that the model is better than halt95/Swift1.5-Qwen3.8-Flash-Next-W4A16-Merlin.