lmstudio-community/Qwen3-8B-MLX-8bit warn
The chat template differs from its base model, which changes behavior.
claims base: Qwen/Qwen3-8B · chat template: present · view on Hugging Face ↗
Scan coverageStatic battery2026-08-21Weights batteryfailedBehavioral batterynot rundetails
| Battery | Looks at | Status |
|---|---|---|
| Static battery | Metadata & packaging | complete 2026-08-21 |
| Weights battery | Weights forensics — no GPU, no download | failedunsupported embedding dtype U32 |
| Behavioral battery | Live-inference differentials | not run |
Ingot runs three batteries against a model. What each one checks →
Findings
Scanned 2026-08-21 · published from a community scan.
medium Chat template differs from claimed parent
The chat template does not match Qwen/Qwen3-8B's. Template drift silently changes model behavior even when weights are identical — 37% of drifted derivatives in our census left it undisclosed. Diff the templates before deploying.
How to fixingot patch
Restore the parent's chat template in `tokenizer_config.json` — a pure metadata fix.
- Run `ingot patch <owner/model>` — the patch manifest carries the parent's template and applies it to a local copy's `tokenizer_config.json`.
- Or fix by hand: copy the `chat_template` value from the parent repo's `tokenizer_config.json` into this model's, and pin your serving stack to that file.
- If the drift was intentional (the author retrained on a new template), confirm that in the model card before "fixing" it — restoring the parent template on retrained weights changes behavior too.
Fix it
Some findings are metadata-level and patchable — apply the fixes to your local copy (your weights never leave your machine):
npx @ingotai/scan patch lmstudio-community/Qwen3-8B-MLX-8bit
Remediation guidance addresses the documented findings only. It is evidence-driven repair, not a safety certification of the model.
Fingerprint
The durable profile of this model: measured weights-and-metadata facts, rebuilt on every scan and battery run. Updated 2026-08-21.
| architecture | qwen3 · 36 layers · 4096-dim |
| parameters | 2303.9M |
| vocabulary | 151,936 tokens |
| license | apache-2.0 |
| serialization | safetensors |
| chat template | present · sha256:5595e8741bb5500f |
| claimed lineage | Qwen/Qwen3-8B |
| lineage verified | unverified — weights battery pending |
Full measured fingerprint
| architectures | Qwen3ForCausalLM |
| library | mlx |
| pipeline | text-generation |
| repo files | 12 |
| revision | fea44fa66d05 |
| HF snapshot | 33.0k downloads · 2 likes · updated 2025-04-28 · captured 2026-08-21 |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/lmstudio-community/Qwen3-8B-MLX-8bit)