LLM EXPLORER 63,636 MODELS INDEXED

Rwkv Raven 3B by RWKV

By RWKV · 72 downloads
HF Repo ↗ Copy URL

Rwkv Raven 3B is an open-source language model by RWKV. Features: 3b LLM, VRAM: 11.9GB, LLM Explorer Score: 0.11, Arc: 36.7, HellaSwag: 59.8, MMLU: 24.9, GSM8K: 0.5.

  Dataset:eleutherai/pile   Endpoints compatible   Pytorch   Region:us   Rwkv   Sharded
Model Card on HF 🤗: https://huggingface.co/RWKV/rwkv-raven-3b 

Rwkv Raven 3B Parameters and Internals

Model Type 
RNN, Transformer
Additional Notes 
RWKV is an RNN with transformer-level LLM performance. It is designed for fast inference, VRAM efficiency, and 'infinite' context length. The model allows for free sentence embedding and fast training.
Input Output 
Input Format:
text
Accepted Modalities:
text
Output Format:
text
Performance Tips:
The 'Raven' models need to be prompted in a specific way. Learn more about that in the integration blog post.
LLM NameRwkv Raven 3B
Repository 🤗https://huggingface.co/RWKV/rwkv-raven-3b 
Model Size3b
Required VRAM11.9 GB
Updated2026-08-07
MaintainerRWKV
Model Typerwkv
Model Files  2.0 GB: 1-of-7   2.0 GB: 2-of-7   2.0 GB: 3-of-7   1.9 GB: 4-of-7   1.9 GB: 5-of-7   1.6 GB: 6-of-7   0.5 GB: 7-of-7
Model ArchitectureRwkvForCausalLM
Transformers Version4.29.0.dev0
Tokenizer ClassGPTNeoXTokenizer
Vocabulary Size50277
Torch Data Typefloat32

Best Alternatives to Rwkv Raven 3B

Best Alternatives
Context / RAM
Downloads
Likes
Hermes RWKV V4 3B0K / 6 GB100
Rwkv 4 3B Pile0K / 11.9 GB1262
Rwkv 5 V2 3B 16K0K / 6.1 GB201
Note: green Score (e.g. "73.2") means that the model is better than RWKV/rwkv-raven-3b.