QinEmPeRoR93/nassila-sanad-9b

<div align="center">

Nassila Sanad 9B

Full-text claim grounding · FT-6 · sole Sanad tier

   

English full-text only · ship verdicts come from the verifier, not the model

</div>

What this is

Local GGUF for Sanad in Nassila: given a manuscript passage and source excerpts, the model returns structured JSON (claims, verdicts, verbatim quotes). The app’s deterministic verifier always owns the final ship decision.

Retired abstract-era siblings (legacy only): `12b` · `e4b`

Download

Start here: `nassila-sanad-9b-q4_k_m.gguf` — best default for LM Studio / Nassila.

Two families, same FT-6 trunk:

Default files were converted with llama.cpp `--no-mtp` (no phantom MTP block). MTP files keep the real Qwen3.5 draft head — prefer a recent llama.cpp; LM Studio may not load them.

Default (no MTP)

MTP (speculative)

Tip: Nassila / LM Studio → default Q4_K_M. llama-server + MTP → mtp-Q4_K_M or mtp-Q6_K.

Quick start

Nassila (recommended)

1.8.0+ · Settings → Passage grounding → runner (LM Studio / Ollama / vLLM / Custom) → model nassila-sanad-9b.

Qwen3.5 thinks by default. Nassila handles that via no-thinking template guidance, max_tokens: 2048, and in-app stripQwenThinkingTraces.

One-liner (llama-server)

llama-server \
  -m nassila-sanad-9b-q4_k_m.gguf \
  --host 127.0.0.1 --port 1234 \
  --ctx-size 8192 --n-gpu-layers 99 \
  --jinja \
  --chat-template-file qwen3.5-no-thinking.jinja

<details> <summary><b>Chat template</b> — <code>qwen3.5-no-thinking.jinja</code> (required for clean JSON)</summary>

Without this, the model emits a thinking trace first and the JSON often truncates.

{% set enable_thinking = false %}
{%- for message in messages %}
{%- if message['role'] == 'system' %}
{{- '<|im_start|>system\n' + message['content'] + '<|im_end|>\n' }}
{%- elif message['role'] == 'user' %}
{{- '<|im_start|>user\n' + message['content'] + '<|im_end|>\n' }}
{%- elif message['role'] == 'assistant' %}
{{- '<|im_start|>assistant\n' + message['content'] + '<|im_end|>\n' }}
{%- endif %}
{%- endfor %}
{%- if add_generation_prompt %}
{{- '<|im_start|>assistant\n' }}
{%- endif %}

Also documented in the Nassila Sanad setup guide (llama.cpp tab).

</details>

<details> <summary><b>Ollama</b></summary>

Needs Ollama 0.5+ and a public Hub repo:

ollama pull huggingface.co/QinEmPeRoR93/nassila-sanad-9b:Q4_K_M

If output truncates, add a Modelfile TEMPLATE that disables thinking, or use the llama-server path above.

</details>

<details> <summary><b>MTP serve</b> (<code>llama-server</code>)</summary>

llama-server \
  -m nassila-sanad-9b-mtp-q4_k_m.gguf \
  --host 127.0.0.1 --port 1234 \
  --ctx-size 8192 --n-gpu-layers 99 \
  --spec-type draft-mtp \
  --jinja \
  --chat-template-file qwen3.5-no-thinking.jinja

</details>

Output shape

Single JSON object, e.g.:

{
  "claims": [
    {
      "claim": "...",
      "verdict": "supported",
      "sourceQuotes": ["..."],
      "rationale": ["..."],
      "hasNumericClaim": false
    }
  ],
  "overallVerdict": "support",
  "overallRationale": ["..."]
}

Eval (FT-6 · seeds 42 / 43 / 44)

Means across three seeds. Soft vs written ≥0.99 bars noted where relevant.

<details> <summary><b>v2 detail</b> — per-verdict accuracy</summary>

Soft vs ≥0.99: parse 0.96, quote 0.988.

</details>

<details> <summary><b>Compound detail</b> — multi-claim structure gates</summary>

Structure gates (split / exact / omit / bundle / overall ≥ 0.90) pass. Treat compound false_supported and quote as soft — validate on private manuscripts.

</details>

Related artifacts

<details> <summary><b>Limitations</b></summary>

• —Advisory grounding only — ship verdicts always come from the deterministic verifier.
• —v2 parse / quote sit slightly under the written ≥0.99 bars; compound false_supported ~0.10.
• —Requires thinking off (enable_thinking=false or Nassila 1.8.0+ handling).
• —Not bundled in the Nassila installer.
• —English-only train/eval; Arabic L3 is unvalidated.

</details>

<details> <summary><b>Base model & license</b></summary>

• —Base: `Qwen/Qwen3.5-9B`
• —Fine-tune weights: Apache-2.0
• —Base model subject to the Qwen license

</details>

<div align="center">

Sanad grounds · Nassila verifies

</div>