For the past few years, Emerge has tracked thousands of mobile apps to understand their size changes.
We were curious about trends in .mlmodel files, so we pulled some data and this is what we found 🧵
We were curious about trends in .mlmodel files, so we pulled some data and this is what we found 🧵
April 10, 2025 at 6:42 PM
For the past few years, Emerge has tracked thousands of mobile apps to understand their size changes.
We were curious about trends in .mlmodel files, so we pulled some data and this is what we found 🧵
We were curious about trends in .mlmodel files, so we pulled some data and this is what we found 🧵
🚨 New app drop!
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore
MagicaLCore offers everything you need to work with machine learning on your iPad.
From importing and organizing models to training and live testing, MagicaLCore lets you build and experiment with m...
apps.apple.com
June 23, 2025 at 3:34 PM
🚨 New app drop!
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
MagicaLCore for iPad lets you create image classifiers using your own dataset — no coding required.
• Build & test CoreML models on-device
• Export .mlmodel in seconds
• Free to try, with 3-day trial
🔗 apps.apple.com/app/id646644...
#indieDev #CoreML #AI #NoCode
#OpenSource #GPT3 #ChatGPT #MLModel #AI #Python
March 9, 2026 at 4:56 PM
#OpenSource #GPT3 #ChatGPT #MLModel #AI #Python
🚗 My new ML course is coming together! You'll learn to train a car price prediction model with scikit-learn, convert it to .mlmodel using coremltools, host it on a Vapor server, and connect it to an iOS app via a JSON API.
May 5, 2025 at 12:49 PM
🚗 My new ML course is coming together! You'll learn to train a car price prediction model with scikit-learn, convert it to .mlmodel using coremltools, host it on a Vapor server, and connect it to an iOS app via a JSON API.
AndroidにMLModelのBindingの設定とかあるのか
```
buildFeatures {
mlModelBinding = true
}
```
```
buildFeatures {
mlModelBinding = true
}
```
February 22, 2024 at 3:55 AM
AndroidにMLModelのBindingの設定とかあるのか
```
buildFeatures {
mlModelBinding = true
}
```
```
buildFeatures {
mlModelBinding = true
}
```
My new course is coming along nicely!
Python + scikit-learn + pandas to train the model. Convert to .mlmodel using coremltools, host it on Vapor and then interact with the model through JSON API through a SwiftUI app.
😎
Hoping to launch next week.
#iosdev #swiftui #MachineLearning
Python + scikit-learn + pandas to train the model. Convert to .mlmodel using coremltools, host it on Vapor and then interact with the model through JSON API through a SwiftUI app.
😎
Hoping to launch next week.
#iosdev #swiftui #MachineLearning
May 8, 2025 at 11:39 PM
My new course is coming along nicely!
Python + scikit-learn + pandas to train the model. Convert to .mlmodel using coremltools, host it on Vapor and then interact with the model through JSON API through a SwiftUI app.
😎
Hoping to launch next week.
#iosdev #swiftui #MachineLearning
Python + scikit-learn + pandas to train the model. Convert to .mlmodel using coremltools, host it on Vapor and then interact with the model through JSON API through a SwiftUI app.
😎
Hoping to launch next week.
#iosdev #swiftui #MachineLearning
CVE-2026-96804 - CVE-2026-96804
CVE ID : CVE-2026-96804
Published : Sept. 23, 2026, 5:17 p.m. | 43 minutes ago
Description : MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model...
CVE ID : CVE-2026-96804
Published : Sept. 23, 2026, 5:17 p.m. | 43 minutes ago
Description : MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model...
CVE-2026-96804 - CVE-2026-96804
MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model(), which allows a remote attacker to execute arbitrary code via a crafted MLmodel artifact.
cvefeed.io
September 23, 2026 at 6:43 PM
CVE-2026-96804 - CVE-2026-96804
CVE ID : CVE-2026-96804
Published : Sept. 23, 2026, 5:17 p.m. | 43 minutes ago
Description : MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model...
CVE ID : CVE-2026-96804
Published : Sept. 23, 2026, 5:17 p.m. | 43 minutes ago
Description : MLflow's statsmodel flavor, versions 2.1.0 to 3.14.0, omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False security control entirely in _load_model...
using ai to mass edit code that I made organically, by hand with the grace of god, just because they told me to rush this.
fucking life
fucking life
May 28, 2026 at 8:59 PM
using ai to mass edit code that I made organically, by hand with the grace of god, just because they told me to rush this.
fucking life
fucking life
MLflow advisory GHSA-gqvg-gmmx-x4hm: the mlflow.statsmodels flavor omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION guard, so setting it to False does not stop pickle execution on that path. Rated high; it needs a crafted MLmodel artifact in a path MLflow loads. #software
September 1, 2026 at 6:29 PM
MLflow advisory GHSA-gqvg-gmmx-x4hm: the mlflow.statsmodels flavor omits the MLFLOW_ALLOW_PICKLE_DESERIALIZATION guard, so setting it to False does not stop pickle execution on that path. Rated high; it needs a crafted MLmodel artifact in a path MLflow loads. #software
📊🤖 Finished building my first linear regression model to predict NY Taxi ride durations using the January-February 2023 dataset. Next step: tuning the model and deploying with Docker. Let's do this! #MLModel #DataScience #AI #ZoomCamp #MLOps #DataTalksClub
May 22, 2025 at 12:03 AM
📊🤖 Finished building my first linear regression model to predict NY Taxi ride durations using the January-February 2023 dataset. Next step: tuning the model and deploying with Docker. Let's do this! #MLModel #DataScience #AI #ZoomCamp #MLOps #DataTalksClub
🐍Python libraries that implement agnostic global explainability methods 👇
#python #machinelearning #MLModel #datascience #dataengineering
#python #machinelearning #MLModel #datascience #dataengineering
August 6, 2025 at 4:02 PM
🐍Python libraries that implement agnostic global explainability methods 👇
#python #machinelearning #MLModel #datascience #dataengineering
#python #machinelearning #MLModel #datascience #dataengineering
Strategic pruning and quantization shrink a 7‑billion‑parameter LLAVA model to run within 4 GB VRAM, cutting memory use by ~70% and delivering a 4% performance gain. Read more: https://getnews.me/efficient-compression-techniques-boost-medical-multimodal-llms/ #mlmodel #compression
September 27, 2025 at 3:19 AM
Strategic pruning and quantization shrink a 7‑billion‑parameter LLAVA model to run within 4 GB VRAM, cutting memory use by ~70% and delivering a 4% performance gain. Read more: https://getnews.me/efficient-compression-techniques-boost-medical-multimodal-llms/ #mlmodel #compression
HiPhO, a new benchmark uses 13 physics Olympiad exams from 2024‑2025 to compare 30 multimodal LLMs with human contestants; closed‑source reasoning models earned 6‑12 gold medals. Read more: https://getnews.me/hipho-benchmark-shows-m-llms-vs-human-physics-winners/ #hipho #physicsolympiad #mlmodel
September 23, 2025 at 1:57 AM
HiPhO, a new benchmark uses 13 physics Olympiad exams from 2024‑2025 to compare 30 multimodal LLMs with human contestants; closed‑source reasoning models earned 6‑12 gold medals. Read more: https://getnews.me/hipho-benchmark-shows-m-llms-vs-human-physics-winners/ #hipho #physicsolympiad #mlmodel
🚨 EUVD-2026-19608
📊 5.1/10
🏢 mlflow
📝 MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated a...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-19608
#cybersecurity #infosec #cve #euvd
📊 5.1/10
🏢 mlflow
📝 MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated a...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-19608
#cybersecurity #infosec #cve #euvd
April 7, 2026 at 2:01 PM
🚨 EUVD-2026-19608
📊 5.1/10
🏢 mlflow
📝 MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated a...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-19608
#cybersecurity #infosec #cve #euvd
📊 5.1/10
🏢 mlflow
📝 MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated a...
🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-19608
#cybersecurity #infosec #cve #euvd
We fetch the bundled model and load it as a MLModel. Then, to enable it to work with the vision framework, we load this as a VNCoreMLModel.
This is important because of the model inputs I mentioned before. The CLIP model only accepts inputs of a specific image resolution and colour space.
June 13, 2025 at 3:02 PM
We fetch the bundled model and load it as a MLModel. Then, to enable it to work with the vision framework, we load this as a VNCoreMLModel.
This is important because of the model inputs I mentioned before. The CLIP model only accepts inputs of a specific image resolution and colour space.
Stored XSS via unsafe YAML parsing in MLflowMLflow is vulnerable to Stored Cr... MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts i...
Origin | Interest | Match
Origin | Interest | Match
CVE-2026-33865 | THREATINT
CVE-2026-33865: MLflow is vulnerable to Stored Cross-Site Scripting (XSS) caused by unsafe parsing of YAML-based MLmodel artifacts in its web interface. An authenticated attacker can upload a malicious MLmodel file containing a payload that executes when another user views the...
cve.threatint.eu
April 7, 2026 at 1:53 PM
The most common ML models used were:
- FindFour.mlmodel (used in 49 apps)
- FourRecognize.mlmodel (49)
- SSDOcr.mlmodel (41)
- DeepLabV3.mlmodel (15)
These are typically used for things like recognizing patterns, scanning documents, and object removal or background editing 🧠
- FindFour.mlmodel (used in 49 apps)
- FourRecognize.mlmodel (49)
- SSDOcr.mlmodel (41)
- DeepLabV3.mlmodel (15)
These are typically used for things like recognizing patterns, scanning documents, and object removal or background editing 🧠
April 10, 2025 at 6:42 PM
The most common ML models used were:
- FindFour.mlmodel (used in 49 apps)
- FourRecognize.mlmodel (49)
- SSDOcr.mlmodel (41)
- DeepLabV3.mlmodel (15)
These are typically used for things like recognizing patterns, scanning documents, and object removal or background editing 🧠
- FindFour.mlmodel (used in 49 apps)
- FourRecognize.mlmodel (49)
- SSDOcr.mlmodel (41)
- DeepLabV3.mlmodel (15)
These are typically used for things like recognizing patterns, scanning documents, and object removal or background editing 🧠
The app with the largest amount of .mlmodel files is the Filmmaker Pro app with 524 MB of ml model files 🫨
The entire app itself is 1 GB so we'll have to save that breakdown for a thread of its own 🫣
The entire app itself is 1 GB so we'll have to save that breakdown for a thread of its own 🫣
April 10, 2025 at 6:42 PM
The app with the largest amount of .mlmodel files is the Filmmaker Pro app with 524 MB of ml model files 🫨
The entire app itself is 1 GB so we'll have to save that breakdown for a thread of its own 🫣
The entire app itself is 1 GB so we'll have to save that breakdown for a thread of its own 🫣
Since 2023, 464 iOS apps shipped with .mlmodel files (of the ones we tracked)
⚖️ Average model size: 20 MB
💥 Largest size: 524 MB
#️⃣ Most apps had 1 .mlmodel per app, but some had more (up to 51!)
⚖️ Average model size: 20 MB
💥 Largest size: 524 MB
#️⃣ Most apps had 1 .mlmodel per app, but some had more (up to 51!)
April 10, 2025 at 6:42 PM
Since 2023, 464 iOS apps shipped with .mlmodel files (of the ones we tracked)
⚖️ Average model size: 20 MB
💥 Largest size: 524 MB
#️⃣ Most apps had 1 .mlmodel per app, but some had more (up to 51!)
⚖️ Average model size: 20 MB
💥 Largest size: 524 MB
#️⃣ Most apps had 1 .mlmodel per app, but some had more (up to 51!)