LLM EXPLORER 59,598 MODELS INDEXED

Mamba2 Primed HQwen3 8B Instruct by amazon

By amazon · 227 downloads
HF Repo ↗ Copy URL

Mamba2 Primed HQwen3 8B Instruct is an open-source language model by amazon. Features: 8b LLM, VRAM: 17GB, Context: 128K, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.24.

  Arxiv:2405.21060   Arxiv:2502.17605 Base model:finetune:qwen/qwen3...   Base model:qwen/qwen3-8b   Conversational   Endpoints compatible   Hybrid   Hybrid qwen3   Instruct   Instruction-tuned   Linear-attention   Long-context   Mamba-2   Priming   Region:us   Safetensors   Sharded   Ssm   State-space-model   Tensorflow
Model Card on HF 🤗: https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct 

Mamba2 Primed HQwen3 8B Instruct Parameters and Internals

LLM NameMamba2 Primed HQwen3 8B Instruct
Repository 🤗https://huggingface.co/amazon/Mamba2-primed-HQwen3-8B-Instruct 
Base Model(s)  Qwen3 8B   Qwen/Qwen3-8B
Model Size8b
Required VRAM17 GB
Updated2026-05-25
Maintaineramazon
Model Typehybrid_qwen3
Instruction-BasedYes
Model Files  4.9 GB: 1-of-4   4.9 GB: 2-of-4   5.0 GB: 3-of-4   2.2 GB: 4-of-4
Model ArchitectureHybridQwen3ForCausalLM
Licenseapache-2.0
Context Length131072
Model Max Length131072
Transformers Version5.3.0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size151936
Errorsreplace

Best Alternatives to Mamba2 Primed HQwen3 8B Instruct

Best Alternatives
Context / RAM
Downloads
Likes
GKA Primed HQwen3 8B Instruct128K / 17 GB73452
GDN Primed HQwen3 8B Instruct128K / 17 GB2792
...OJOF Primed HQwen3 8B Instruct128K / 18.5 GB2212
Note: green Score (e.g. "73.2") means that the model is better than amazon/Mamba2-primed-HQwen3-8B-Instruct.