---
title: Blockway/Agens-Volundr-32B-Preview
url: https://recentmodel.com/models/Blockway/Agens-Volundr-32B-Preview/
status: archived
published: 2026-10-03T09:04:47Z
updated: 2026-10-19T23:17:53Z
source: https://huggingface.co/Blockway/Agens-Volundr-32B-Preview
---

# Blockway/Agens-Volundr-32B-Preview

Agens Volundr 32B is an open-weights model from Blockway, built for reasoning, coding and agent work. Dense, 33B params. Context: 256K tokens (config). License: apache-2.0.


- **Org:** Blockway
- **Released:** 2026-10-03
- **Status:** archived
- **License:** apache-2.0 (permissive)
- **Parameters:** 32.7B
- **Architecture:** dense · VolundrForConditionalGeneration
- **Context:** 262,144 tokens
- **Task:** text-generation · text
- **Signal:** 153 HF likes · 804 downloads
- **Archived:** 2026-10-19 (no longer listed by the collector; stats as last seen)

## Runs on

real quant files; sizing calculation, not tested.

- Q4 ~30.0 GB: 32GB GPU (5090), Mac 48GB
- Q8 ~39.7 GB: 48GB GPU (RTX 6000 Ada / A6000), Mac 64GB (M1 Max), CPU only, 64GB RAM (slow)
- FP16 ~67.9 GB: 96GB GPU (RTX Pro 6000 Blackwell), 1× DGX Spark (128GB), 2× DGX Spark (256GB), Strix Halo 128GB unified (~96GB usable), Mac 96GB, Mac 128GB, Mac 192GB, Mac 256GB, Mac 512GB, CPU only, 128GB RAM (slow), CPU only, 256GB RAM (slow)

## Links

- [Hugging Face](https://huggingface.co/Blockway/Agens-Volundr-32B-Preview)
- [GGUF: Blockway/Agens-Volundr-32B-Preview-GGUF](https://huggingface.co/Blockway/Agens-Volundr-32B-Preview-GGUF)