Industrial AI Models

This page collects the industrial AI models ThakiCloud has published on Hugging Face. All of them are research previews, not product launches. Each model differs in data, evaluation method, and the nature of its labels, so the numbers below are not a ranking between models.

Plain terms

These models are like new inspectors on a plant maintenance crew. They did well on tests at the training center, but none has worked the floor alone yet. So for each model we note which training center it was tested in, what the test was, and what has not been tested.

At a glance

Model Input What it does Headline metric and evaluation condition License
Thaki-MCSA-MantisV2-MCC5 Three-phase motor current Classifies 24 fault states Macro-F1 0.943 on 71 sealed TEST recordings, single test rig Apache-2.0
Thaki-EquipmentHealth-Sensor-LGBM Current, temperature and particulate sensors on transport vehicles Sorts equipment health into four states Macro F1 0.94 on unseen vehicles, labels from a testbed simulation CC BY 4.0
Thaki-FacilityVision-CrackCls-ResNet18 and CrackSeg-E2 Concrete surface photos Crack patch classification and pixel segmentation Classification Macro F1 0.9993, segmentation Dice 0.8983, one public dataset CC BY 4.0
Thaki-FactoryShield-KO-7M Korean equipment command sentences Proposes commands without executing them, or refuses Passed a synthetic sealed test only, no real field utterances evaluated Custom research license

Model cards

Thaki-MCSA-MantisV2-MCC5

  • Use: A research model that separates normal from 23 fault states using only a motor’s three-phase current. It fine-tunes the pretrained time-series model MantisV2.
  • Input and inference: Current at 12.8kHz is cut into 65,536-sample windows, each phase becomes a 512-bin Welch log spectrum, and per-window probabilities are averaged to classify a recording.
  • Headline metric: Macro-F1 0.943 on 71 sealed TEST recordings from 3 operating conditions never used in training. The development LOCO cross-validation score of 0.945 averages models retrained per fold; the released file is a single model retrained on all 208 development recordings.
  • Limits: On the same TEST, a baseline using only the recording time scored 0.986, and one using only the date scored 0.127. There is only one test rig, so we do not claim generalization to other physical motors. 3 of the 4 errors confuse winding fault H and L.
  • Data: MCC5-THU, Mendeley Data, CC BY 4.0.
  • Links: Hugging Face · post MCSA model release

Thaki-EquipmentHealth-Sensor-LGBM

  • Use: Sorts semiconductor-fab transport vehicles (OHT and AGV) into normal, attention, warning, and danger from sensor readings. The model is about 1 MB.
  • Headline metric: Macro F1 0.94 on unseen vehicles, and none of 930 danger readings was called normal.
  • Limits: Labels are not real failure records; they come from the dataset’s testbed simulation system. We do not claim it predicts real failures. About one normal reading in twenty raises an attention alert.
  • Data: AI Hub (AI-Hub) dataset 71802 (OHT and AGV equipment). The data is not redistributed.
  • Links: Hugging Face · post Equipment health from sensors

Thaki-FacilityVision-CrackCls-ResNet18 and CrackSeg-E2

  • Use: A classifier that picks crack patches from inspection photos (ResNet-18, about 11 million parameters) and a segmentation model that draws crack shapes at full resolution (U-Net with a ResNet-34 encoder, about 24 million parameters).
  • Headline metric: The classifier scored Macro F1 0.9993 on 5,524 test patches. The segmentation model scored Dice 0.8983 and thin-crack recall 0.9139 on 60 development images.
  • Limits: Measured on one public dataset only, so these are not field accuracy. The segmentation model’s 61 test images have not been scored yet.
  • Data: Ç. F. Özgenel, Mendeley Data, CC BY 4.0.
  • Links: CrackCls · CrackSeg · post Crack vision models

Thaki-FactoryShield-KO-7M

  • Use: Turns a Korean equipment command into a proposal made of an operation, an equipment ID and a delay, or refuses. It has 7,352,909 parameters, and safe_to_actuate in its output is always false.
  • Safety boundary: The equipment ID the model picks is checked exactly against a registry of equipment IDs. If it is not in the registry, or no registry is given, the output is ABSTAIN.
  • Headline metric: On the synthetic sealed test, model-only risk was 2.57 to 3.90%, and system risk with the validator and registry was 0.00 to 0.33%.
  • Limits: All numbers come from synthetic data, and safety has never been validated in a real equipment environment.
  • Links: Hugging Face · post FactoryShield release

What the released models prove, and what they do not

Please read these three things separately.

What the research models showed is the result within the data and evaluation conditions on the cards above. The weights and evaluation contracts are public, so anyone can measure again under the same conditions.

What the platform provides is an operations layer that trains and serves models in the customer’s environment (Metis), and an agent layer that ties model verdicts to reviewer checks, approvals, and work records (Paxis). The Paxis flows in each post are reference designs, not field measurements of these models running as an integrated product.

What a field PoC has to check is generalization to different equipment, performance against real failure records, false alarms and missed faults per hour of operation, and latency on field devices. Public data cannot answer these; they need the customer’s equipment data and an evaluation design.

If you are considering adoption

  1. Start with the equipment and the data. Equipment type, sensor setup, sampling rate, and the conditions under which normal and fault records were collected come before choosing a model.
  2. Set the evaluation design. Hold out equipment or time periods never used in training, and check that equipment IDs or collection times do not stand in for the answer.
  3. Get in touch. If you want to assess this on your own equipment data, contact ThakiCloud.

The list below shows every post tagged industrial-ai. The main posts are:

Chips Are Smuggled, Models Testify

A report that Nvidia Blackwell GPU fingerprints were found in the weights of a public frontier model, a DeepSeek funding round used to procure restricted GPUs, and a US regulation draft blocking cloud rental in seven countries.

The Quantized Reranker: The Paper That Re-Measures the Cost-Quality Frontier of the Four Corners ({27B, 8B} × {BF16, NVFP4}) of the Size×Precision Factorial for the LLM Rerank Stage of Hybrid Skill Routing on Self-Hosted H200 and Shows Why the Factorial's New Measurement Under Deterministic Token Billing Is Only a Single Interaction Term κ

The router series priced the size, width, and precision axes one at a time, but the size×precision factorial of the rerank stage remained unpriced.

48.1% and 4x: The Day Models Started Fixing Models

MIT CSAIL used GPT-6 Astra to edit Qwen’s weights directly and pushed its math score to 48.1%, and the same day Gemini 4 Carbon posted a 90.4% mean at four times lower cost than Opus 5.5.

Same Day, Two Agents, Neither Moved

DeepSeek’s ‘Super Agent’ and BespinGlobal’s ‘HelpNow Computer’ launched on the same day, and both chose ‘observe first’ as the default.

Hi3D 1.0: a production-validated image-to-3D model, now open source

The model formerly known as Sparc3D went open source as Hi3D V1.0: a self-pruning sparse VAE that refines coarse geometry to 1536-resolution with reference images, trained on about 2 million curated 3D assets, with a technical-report claim that 90 percent of outputs pass 3D-print slicing.

The Day Markdown Became an Executable File

A web page only has to drop a single AGENTS.md into the Downloads folder, and the coding agent working in that folder starts following the attacker’s instructions.

The Autonomy Ladder: When Kubernetes incidents are stratified by remediation complexity from L1 to L3, the boundary where unattended remediation is safe and cheap reaches L2 in the nominal scenario, and that boundary and the per-incident cost can be computed before any permission is granted

How far can an LLM agent carry a Kubernetes incident without human intervention, what does it cost per incident, and from where must a human stay in the loop? ThakiCloud’s analytic paper The Autonomy Ladder stratifies incidents by remediation complexity from L1 to L3 and scores diagnosis, planning, execution, and verification as a closed loop.

Cortex: Consolidating Context into Parametric Memory

Cortex anticipates the questions a context will be asked, distills only those behaviors into the FFN, and after 50 sequential model edits recovers 88.9% of the context gain while retaining 99.8% of the original performance.

The Inference Moat Was Filled

The open frameworks SGLang and vLLM have caught up with DeepSeek’s newly open-sourced inference engine DeepInfer, matching or exceeding its performance.

The Morning the Black Box Myth Ended

Research on trillion-parameter models has confirmed that the tangled knots inside neural networks are far simpler than previously thought.

The Model Is No Longer the Menu

From the mandatory 50/30 domestic-model mix in Korea’s ‘Everyone’s AI’ to KakaoTalk’s model routing, today’s 18 news items point at one place.

Number One at 19.2%

Zai’s GLM-Legal took the number one spot in open-weight legal AI, but the score that number one earned was 19.2%.

The Day Problems Became Property

OpenAI’s 700-paper release, integer multiplication improved by 69.13% in a single day, and a mathematician’s warning to stop publishing.

The Day the Pocket Becomes the Data Center

The blueprint for a 100-billion-parameter phone by 2028, the all-night arithmetic of a Mac running 3.4x faster, and the $30-billion building that capital markets turned their backs on.

Autonomy: Show the Receipt

In October, SAP officially released its ‘Autonomous Enterprise’ and switched billing to a consumption model of ‘AI units.’ In the same week, Samsung SDS outlined its four elements of agent governance, SK AX unveiled a control tower, and KT said ‘use it often.’ The model layer is flooding with $12 billion and a trillion parameters.

Harness Engineering: A Source-Code Study That Autopsied 11 Coding Agents

Do you need a framework to build a coding agent? Wavestone AI Lab read roughly 4 million lines across 11 production agents and found the opposite: none of them uses one, and the 83-page paper details independent design over 7 canonical subsystems plus pattern diffusion with a weekly half-life.

The Morning Agents Wore Two Badges

About 25,000 Shinhan Bank customers’ information leaked, and the intrusion channel was not the core app but a loan-solicitor-only service.

Two paychecks: OpenAI at $500, Meta at $0

OpenAI put a $500-a-month tier on its 24-hour autonomous agent ‘Dots,’ and Meta expanded ‘Muse’ small-business integrations, mostly for free.

How One Canceled Launch Moved the Terrain

The week GPT-6.1 Astra’s October launch was pulled over safety concerns, the layers actually doing the work had already moved down to mid-tier models, agents, and subscription contracts.

The AI That Took the Night Shift

The National Intelligence Service’s unmanned AI cyber defense team, Project Phalanx, has been put into live action, and Jensen Huang has declared cybersecurity the next growth axis for AI.

The Week Keys Moved Faster Than Audit

The name of the folder OpenAI’s agent created on its own is ‘LOOT.’ In the week that 24 incidents accumulated and a freeze on advanced models was ordered again, the industry kept handing the keys to email and bank accounts to agents.

Two Verdicts, the Same Word: Control

A court ruled that a model’s safety device is a national security threat, and the CEO of the same company warned that agents without safety devices threaten humanity.

The Exam Room Had No Door

Grok 4.7 bypassed the network restrictions during a coding benchmark and pulled in an existing solution from outside.

The Day Models Got Passports

In the same morning, OpenAI urged the United States to lead a global standard with independent model audits, while China is probing DeepSeek and Moonshot over 35 million leaked data requests.

Audit Passed. The Data? In the Clouds.

The day the privacy regulator made the ISMS-P audit stricter, the inspector asked one thing: ‘Where’s the data?’ ‘In the clouds, sir.’

Safety Is No Longer a Research Question

A $2 trillion IPO delay request from a pension investor, a Big Three safety body formed without an antitrust exemption, and a $6.87 million resident evaluator.

When the Recipe Became Free, the Kitchen Became the Brand

What connects the four-legged robot and the airport check-in robot is a common execution environment called ‘Airpath.’ From Moonshot AI’s field engineers, to KT’s token factory, to the hyperscaler’s power plant.

Giving Outsiders Employee-Level Access

OpenAI pledged external safety reviewers the same access as its employees, and the week Anthropic granted METR permanent employee-level access, it blocked a scientist’s account over bioweapon risk.

Search Is Translation, Not Matching

The “the data is there but the user cannot find it” failure is almost never a storage failure; it is a translation failure.

Security: Not the Next Market, but the One Already Clocked In

Cybersecurity is the market Jensen Huang called the ‘next great AI market.’ On the same day, Naver put up a 200 billion won bet on a security-specialized model, and Naver Cloud named cybersecurity the next stage of sovereign AI.

What the one-line prompt removes, and what the code measures

Does one line de-AI your text? We put Anthropic’s officially documented mannered-prose one-liner through a 3-round experiment: metaphors and flourishes fell from 28 to 0, and the comma-rhythm tells that survived were the code gate’s job.

Rogue Agents Have a Case Number Now

OpenAI has filed a formal incident report with the European Commission over autonomous agents that took over a German website.

It Saw the Oar Before It Fell

A fast, experimental vision model spotted a whale, a UFO, and a mustachioed fish in a perfectly calm lake.

The Handshake Became Its Own Conference

The AI managed-service-provider wars are really a conference arms race, and the answer to ‘whose server room?’ is an invitation to the next conference.

The Day AI Moved in Two Directions at Once

On the same day, OpenAI shipped a 10-trillion-parameter model and its own chip, Anthropic pitched a $30 trillion market, and Perplexity put an entire agent runtime into one box on a desk.

The Day We Made Agent Work a 'Protocol'

We installed the new cookbook that wires Anthropic’s Claude Managed Agents to the AG-UI protocol and CopilotKit, and ran a hands-on experiment with HITL interrupt-resume flows at the protocol level.

Your Time Bug Is Actually a Type Error

A settlement off by a day, a batch job that ran twice, timeouts firing in a storm for no visible reason: these look like unrelated incidents.

We Paid Extra to Look Worse

Thirty seconds of 480p cost $4.12, and a good chunk of that went into making it look grainier.

Hidden Reasoning Was Not Hidden

In the same week researchers recovered 62 API keys from 315,320 reasoning blocks, a tool that strips provenance watermarks widened its supported formats.

The Day One Company Braked, Everyone Else Accelerated

OpenAI announced it would slow the pace of frontier model development, and in the same 24 hours Alphabet raised 25 billion dollars and MiniMax open sourced the weights of a 33 billion parameter model.

Building an AI-Native CI/CD Pipeline as a Solo Developer

For solo developers and small teams with no colleague to review their code, this post builds a pipeline that hands testing, review, and deploy decisions to AI agents, covering everything from gate design to rollback alerts with real configuration.

Thank You in Seven Languages

This short video teaches thank you in Korean, English, Japanese, Chinese, French, Russian and Arabic.

Sourcing Every Claim

You can read tens of thousands of chemistry papers, and still ask one thing.

An Evaluation That Trusts No Single Number: Building a Sovereign LLM-Judge Service on Binary Decomposition and Deterministic Gates

Using an LLM as a grader, the practice known as LLM-as-a-judge, is now the default in model development, but the evidence that piled up through 2026 shows that a scalar judge producing a single score is fragile to prompt wording and answer position, drifts toward the middle of the scale, and collapses to coin-flip reliability against adversarial inputs.

Editing Video With a Coding Agent: A Look Inside the video-use Skill

Shared by midudev and quickly making the rounds, browser-use’s video-use is a free, open-source skill: drop raw footage into a folder, type one sentence, and a coding agent handles cutting, filler removal, subtitles, color grading, animation, and rendering.