---
base_model: utter-project/EuroLLM-9B-Instruct
license: apache-2.0
quantized_by: tiararodney
library_name: gguf
pipeline_tag: text-generation
tags:
  - gguf
  - llama.cpp
  - eurollm
  - sek
---

# EuroLLM-9B-Instruct GGUF

GGUF quantizations of [utter-project/EuroLLM-9B-Instruct](https://huggingface.co/utter-project/EuroLLM-9B-Instruct),
converted from the original bf16 weights with llama.cpp.

Prepared as a subject for the [sek scrollback-priming cross-model study](https://blog.tiararodney.com/projects/2026/semantic-execution-kernel/experiments/from-seed-to-weights/), where
EuroLLM-9B stands in as the distributionally distant baseline: a multilingual
European-language model rather than an English- and code-centric one. The
question it probes is whether synthetic-scrollback priming can hold a model whose
training mass is natural-language prose in consistent POSIX shell syntax.

## Provenance

- Source: utter-project/EuroLLM-9B-Instruct (bf16 safetensors)
- Converted: llama.cpp convert_hf_to_gguf.py --outtype bf16
- Quantized: llama.cpp llama-quantize
- Built on a single Tesla V100-SXM2-32GB

## Quants

| File | Quant | Size |
| --- | --- | --- |
| filled after the build | | |

## Prompt format

ChatML, with a system role. Stop token: the im-end token.