Qwen/Qwen3-8B warn
claims base: Qwen/Qwen3-8B-Base · chat template: present · view on Hugging Face ↗
Ingot findings
Base model used as census parent; the risk surface is its derivative ecosystem. Weights battery: embedding-norm scan over 151936 tokens (BF16, 4096-dim) found 2999 undertrained candidates, 102 plain-ASCII. Lineage vs Qwen/Qwen3-8B-Base: consistent. Scanned 2026-08-20 (published from a community scan). Full write-up: read the report.
info Derivative ecosystem drift
Across the most-downloaded community derivatives of this model, 56% ship a different chat template than the parent (37% undisclosed), and 78% of pure quantization re-releases changed or dropped the template.
How to fix
Informational — applies to the model's derivative ecosystem, not this checkpoint.
- When choosing a derivative, scan the derivative itself; parent verdicts do not transfer.
info Small weight delta ≠ small behavior delta
The derivative with the smallest measured weight change removed the most safety: harmful-request refusal fell 100% → 5% with zero capability loss. Weight-size scans cannot substitute for behavioral differentials.
How to fix
Informational — applies to the model's derivative ecosystem, not this checkpoint.
- When choosing a derivative, scan the derivative itself; parent verdicts do not transfer.
medium Undertrained (glitch) token surface in vocabulary
Embedding-norm scan flagged 2999 undertrained tokens (norm < 0.3× the vocabulary median of 1.451), including 102 plain-ASCII strings that can appear in ordinary input as identifiers — e.g. "$PostalCodesNL", "ForCanBeConvertedToF", "PostalCodesNL", "ForCanBeConverted", "useRalative", "thuisontvangst", "useRal", "sexkontakte". In models where this class was tested behaviorally, such tokens silently rewrote user input into confident, schema-valid, wrong output. These are candidates from the weights alone; behavioral confirmation requires the GPU deep battery.
How to fixruntime guardweight-level
Keep the affected token strings out of the model's input — the scan-derived runtime guard carries this model's exact blocklist.
- Fetch this model's guard artifact (`/api/v1/guard/<owner>/<model>`): the confirmed corrupting tokens and the low-norm candidate list, derived from the published scan.
- Screen inbound text with it (the `@ingotai/guard` package is a reference implementation) and route flagged records to a different model or human review — verbatim-copy tasks on flagged strings are the failure mode.
- The underlying cause is undertrained embeddings in the weights; a true fix is weight-level (continued pretraining on the affected tokens) — that is not a patch, it's a training job.
info Weights consistent with claimed parent Qwen/Qwen3-8B-Base
Mean cosine similarity of 64 sampled token-embedding rows against Qwen/Qwen3-8B-Base is 0.998 — the weights plausibly descend from the declared base (relation: unspecified).
How to fix
Fix or verify the `base_model` declaration so lineage checks can run.
- If you own the repo: correct the `base_model` field in the model card metadata to the real, public parent.
- If you don't: identify the true parent (config architecture + weight shapes narrow it fast) and re-scan with that lineage in mind.
Remediation guidance addresses the documented findings only. It is evidence-driven repair, not a safety certification of the model.
Fingerprint
The durable weights-and-metadata profile of this model, rebuilt on every scan and deep-battery run. Updated 2026-08-20.
| architecture | qwen3 · 36 layers · 4096-dim |
| parameters | 8190.7M |
| vocabulary | 151,936 tokens |
| license | apache-2.0 |
| serialization | safetensors |
| chat template | present · sha256:a55ee1b1660128b7 |
| claimed lineage | Qwen/Qwen3-8B-Base |
| lineage verified | consistent vs Qwen/Qwen3-8B-Base — embedding-row cosine 0.998 |
| glitch-token surface | 2,999 undertrained candidates, 102 plain-ASCII |
Full fingerprint
| architectures | Qwen3ForCausalLM |
| library | transformers |
| pipeline | text-generation |
| repo files | 15 |
| revision | b968826d9c46 |
| HF snapshot | 15.7M downloads · 1.3k likes · updated 2025-07-26 · captured 2026-08-20 |
| embedding tensor | model.embed_tokens.weight · BF16 · 151,936×4096 |
| embedding norms | median 1.4512 · mean 1.3758 |
| lineage check | consistent — cosine 0.9979 over 64 sampled rows vs Qwen/Qwen3-8B-Base |
| glitch-token samples | "$PostalCodesNL", "ForCanBeConvertedToF", "PostalCodesNL", "ForCanBeConverted", "useRalative", "thuisontvangst", "useRal", "sexkontakte", "NdrFc", "webElementX", "sextreffen", "wannonce" |
Verdict badge
Ship the verdict in your README — it always shows the latest published analysis:
[](https://ingot.tools/models/Qwen/Qwen3-8B)