Model page

deepseek-ai/DeepSeek-OCR warn

downloads 2.3Mlikes 3.3klicense mitarch deepseek_vl_v2params 3336.1Mupdated 2025-11-04

chat template: not found · view on Hugging Face ↗

Scan coverage

Ingot runs three batteries against a model. What each one checks →

BatteryLooks atStatus
Static batteryMetadata & packagingcomplete 2026-08-22
Weights batteryWeights forensics — no GPU, no downloadcomplete 2026-08-22129,280-token embedding scanned · 829 undertrained
Behavioral batteryLive-inference differentialsnot run

Findings

Scanned 2026-08-22 · published from a community scan.

medium Repo ships executable Python (trust_remote_code)

The repository contains custom code files that run in-process when loaded with trust_remote_code=True. Pin the revision hash and review the code before loading.

How to fix

Review and pin the custom code; never float on `main` with trust_remote_code=True.

  1. Read every `.py` file in the repo before first load — this code runs in your process.
  2. Pin the revision: `from_pretrained(model_id, revision="<commit sha>", trust_remote_code=True)` so a later push can't swap the code under you.
  3. Prefer a version of the architecture already in `transformers` if one exists, which removes the remote-code requirement entirely.

medium Undertrained (glitch) token surface in vocabulary

Embedding-norm scan flagged 829 undertrained tokens (norm < 0.3× the vocabulary median of 7.711), including 172 plain-ASCII strings that can appear in ordinary input as identifiers — e.g. "DelaLika", "Kadaghanon", "Pagbuok", "Kapunoang", "Kasipak", "Kasarangang", "Pagklas", "Kaginharian". 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 behavioral 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.

  1. 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.
  2. 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.
  3. 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.

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 2026-08-22.

architecturedeepseek_vl_v2 · 12 layers · 1280-dim
parameters3336.1M
vocabulary129,280 tokens
licensemit
serializationsafetensors custom code
chat templatenone
glitch-token surface829 undertrained candidates, 172 plain-ASCII
Full measured fingerprint
architecturesDeepseekOCRForCausalLM
librarytransformers
pipelineimage-text-to-text
repo files21
revision9f30c71f441d
HF snapshot2.3M downloads · 3.3k likes · updated 2025-11-04 · captured 2026-08-21
embedding tensormodel.embed_tokens.weight · BF16 · 129,280×1280
embedding normsmedian 7.7109 · mean 7.5166
lineage checkno claimed base model
glitch-token samples"DelaLika", "Kadaghanon", "Pagbuok", "Kapunoang", "Kasipak", "Kasarangang", "Pagklas", "Kaginharian", "Beskjeftigelse", "Britonhon", "Kahutong", "kabungtor"

Battery runs

The run trace behind the findings above: every deep-battery job for this model, with what each run measured or why it failed. Findings are only as good as the runs that produced them.

batterystatusqueueddurationattempts
weightscomplete2026-08-21 07:428s1
weights run 2026-08-21 measurements
probes runglitch-norm-scan, lineage-norm-correlation
embedding tensormodel.embed_tokens.weight · BF16 · 129,280×1280
glitch surface829 undertrained, 172 plain-ASCII
lineage checknot checked (no claimed base model)

Verdict badge

Ship the verdict in your README — it always shows the latest published analysis:

Ingot verdict: warn

[![Ingot scan](https://ingot.tools/api/v1/models/deepseek-ai/DeepSeek-OCR/badge.svg)](https://ingot.tools/models/deepseek-ai/DeepSeek-OCR)
Gate it in CI