LLM EXPLORER 61,918 MODELS INDEXED

Maba V2 Architecture by AndrewThompson1233

By AndrewThompson1233 · 609 downloads
HF Repo ↗ Copy URL

Maba V2 Architecture is an open-source language model by AndrewThompson1233. Features: 1m LLM, Context: 4K, License: other, LLM Explorer Score: 0.34.

  100m   1b   1m-context   3-stream   30b   3b   7b   Architecture   Centroid-indexing Decoupled-gated-delta-attentio...   Deepseek   Dg-indexer   Dgda   Efficient-llm   En   Endpoints compatible   Flash-attention   Gated-deltanet   Hca   Linear-attention   Linear-recurrence   Long-context   Maba   Maba-sa   Maba-v2   Maba-v2-architecture   Maba sparse   Mamba   Mamba-2   Minicpm   Mla   Mtp   Multi-head-latent-attention   Multi-token-prediction   Needle-in-a-haystack   Nope   On-device-ai   Pytorch   Qwen   Recurrent   Region:us   Rmsnorm   Rwkv   Scaling   Sparse-attention   Speculative-decoding   Ssm   State-space-model   Sub-quadratic   Swiglu   Triton
Model Card on HF 🤗: https://huggingface.co/AndrewThompson1233/maba-v2-architecture 

Maba V2 Architecture Parameters and Internals

LLM NameMaba V2 Architecture
Repository 🤗https://huggingface.co/AndrewThompson1233/maba-v2-architecture 
Model Size1m
Updated2026-09-22
MaintainerAndrewThompson1233
Model Typemaba_sparse
Supported Languagesen
Model ArchitectureMabaSparseForCausalLM
Licenseother
Context Length4096
Model Max Length4096
Transformers Version4.49.0
Vocabulary Size32768
Torch Data Typefloat32