DeepSeek-R1-Distill-Llama-70B-abliterated

Model source

Published by huihui-ai on Hugging Face. This index organizes public repository metadata for discovery and comparison. Refer to the publisher and independent testing for actual model behavior.

Declared base model

deepseek-ai/DeepSeek-R1-Distill-Llama-70B ↗

Hardware and context

Estimated at 41.3 GB VRAM for 70B parameters with 4-bit weights, including 15% weight overhead and 1 GB of baseline headroom. KV cache, context length, multimodal components and concurrency increase actual memory usage.The parameter count was parsed from the model name and has not been independently verified.

Context length: Not stated in public metadata. Please consult the model card.