---

license: apache-2.0

base_model:

- Blackfrost-AI/Qwythos-9B-V2-EMERGED-BF16

pipeline_tag: text-generation

library_name: gguf

tags:

- reasoning

- agentic

- distillation

- qwen3_5

- multimodal

- vision

- vlm

- en

---

Qwythos-9B-EMERGED — GGUF

GGUF build of Blackfrost-AI/Qwythos-9B-V2-EMERGED-BF16 — the pure-reasoning / agentic distillation of the Qwythos 9B base, trained by Blackfrost AI on The Void. This repository packages those weights in the GGUF format for llama.cpp, Ollama, and LM Studio.

> See the BF16 model card for the full model description, training method, and safety posture. This card documents only the GGUF conversion. The architecture, capabilities, limitations, and responsible-use guidance are identical to the source model and are not repeated here.

---

What this is

| | |

|---|---|

| Source model | Blackfrost-AI/Qwythos-9B-V2-EMERGED-BF16 |

| Format | GGUF (quantized for CPU/edge inference runtimes) |

| Base family | qwen3_5 (Qwen3.6-class) dense 9B, ~9.5B params |

| Modality | Vision-language — text + image + video in, text out (vision tower inherited from base; see source card) |

| Precision | quantized (see per-file list below) |

| Context | 1,048,576 (1M) native, inherited from the base |

This is a format conversion of the BF16 weights — no retraining, no modification of the underlying model. All behavioral characteristics (including limitations and the inherited vision tower) are identical to the source model.

Run (llama.cpp)

# one-shot
llama-cli -m qwythos-9b-emerged.<QUANT>.gguf -p "Explain why a hash table has O(1) average lookup."

# OpenAI-compatible server
llama-server -m qwythos-9b-emerged.<QUANT>.gguf --port 8000

Ollama: ollama create qwythos-9b-emerged -f Modelfile (point FROM at the local GGUF file).

> For multimodal (image/video) input, use a runtime with the source model's vision processor. The GGUF packaging is intended primarily for text inference; validate multimodal behavior against the BF16 source for your use case.

Responsible use

This is an uncensored reasoning model — refusal behavior was removed during distillation. You are responsible for how you use it and for any safety layer you place around it. Do not use it to produce content that is illegal in your jurisdiction. See the source model card for the full safety posture and limitations.

License

Apache-2.0, inherited from the base model lineage (empero-ai/Qwythos-9B-v2).

Provenance

• Source model: Blackfrost-AI/Qwythos-9B-V2-EMERGED-BF16 (Blackfrost AI — The Void distillation)
• Base model: empero-ai/Qwythos-9B-Claude-Mythos-5-1M (empero-ai, Apache-2.0)
• Format tooling: llama.cpp GGUF conversion utilities

Trained and released by Blackfrost AI on The Void. Base model by empero-ai (Apache-2.0).