---

license: other

base_model: unsloth/granite-4.1-8b

base_model_relation: quantized

tags:

- chaosnexus

- chaosnexus-tuned

- gguf

- q4_k_m

- rhai

- mcp

- local-llm

- code

- candle

language:

- en

library_name: gguf

pipeline_tag: text-generation

---

<!-- launch/model/ChaosNexus_Tuned_v1-GGUF/README.md -->

ChaosNexus Tuned v1 (GGUF)

Default ChaosNexus Crucible download. Q4_K_M GGUF of ChaosNexus Tuned v1 (merged Granite 4.1-8B + Tuned LoRA).

| Field | Value |

|-------|-------|

| Hub id | TunedChaos/ChaosNexus_Tuned_v1-GGUF |

| Quant | Q4_K_M |

| Base | unsloth/granite-4.1-8b (IBM Granite 4.1-8B-Instruct) |

| Source adapter | TunedChaos/ChaosNexus_Tuned_v1 |

| Anvil eval mean | 0.944 (smoke CLEAR) - see LoRA card / eval_scores_v1.md |

Intended use

• Local Crucible inference (~/.chaosnexus/crucible/models cache)
• Forge Models tab Pull / Ensure with optional Hugging Face token
• Offline developer machines under ChaosNexus sandbox + human-in-the-loop

How Crucible loads it

1. Set HF token in Forge Settings (Models) if needed for rate limits / gated assets.
2. Pull preset ChaosNexus Tuned v1 (GGUF) (or leave default model_id).
3. Crucible downloads into ~/.chaosnexus/crucible/models and loads via Candle GGUF.

PEFT / LoRA users

Trainers who want the adapter only should use TunedChaos/ChaosNexus_Tuned_v1 instead of this GGUF.

License

• Base model: follow IBM Granite / Unsloth redistributor terms on the Hub.
• Tuned packaging: AGPL-3.0-or-later unless a commercial license is obtained from Tuned Chaos.

Links

• Docs: https://chaosnexus.ai
• Tuned repo: https://codeberg.org/TunedChaos/chaosnexus-tuned
• Project: Tuned Chaos