MLXHub (@MLXHub) header
MLXHub (@MLXHub) profile picture

MLXHub

@MLXHub

Swift-native app for LLM inference on-device using Apple’s MLX. Take full benefit of your Apple hardware and OSS LLMs, your creativity is your limit.

mlxhub.app
Joined July 2026
  • Tweets56
  • Following39
  • Followers23
6Media
Photo by MLXHub (@MLXHub)Photo by MLXHub (@MLXHub)Photo by MLXHub (@MLXHub)Photo by MLXHub (@MLXHub)Photo by MLXHub (@MLXHub)Photo by MLXHub (@MLXHub)
こんにちは日本！@MLXHub アプリがiPhone Duoでめっちゃかっこよく見える！
Image from Juan Colilla (@JuanColilla)
Image from Juan Colilla (@JuanColilla)
1
4
269
BENCHMARK IS HERE! Benchmark your iPhone and iPad with MLXHub against the baseline (M1/8GB). Unlike GeekBench, MLXHub provides an extensive benchmarking to know how well AI models can run in your device, beat the TOP 10! mlxhub.app/ranking/
Device ranking — MLXHub

Device ranking — MLXHub

Top devices in Benchmark v1, the MLXHub benchmark for on-device language models. 1000 = iPad Pro 12.9″ (M1) with 8 GB.

mlxhub.app
1
1
58
Image Generation has arrived! Use MLXHub to generate images locally. No internet required, no limits on the amount. With Plus you can have better resolution!
1
2
54
Voice Center is live! With the 2.3.0 version of MLXHub attach your iPad to a wall (connect it to the power) and you’ll have your LAN Server for IoT and inference, plus a local voice assistant for all your family members!
Image from MLXHub (@MLXHub)
1
1
31
UPDATE: The version 2.3.0 is out! You should notice improvements in the onboarding and the clarity of the setup process, either you want to use it as a Server or as offline ai assistant! Tons of new features and compatibilities added!
1
1
21
UPDATE: MLXHub now supports using Apple Neural Engine to boost MLX performance inside the app, update to version 2.2.1 to enjoy this improvement! apps.apple.com/app/id6766485…
1
1
42
If you are an MLXHub user that you follow us and you see we do not have so much activity, worry not! We are in silent while we are cooking, so expect for it 🤭
32
Congratulations to @Aleph__Alpha for releasing Kolibri-1! Nice to see a new foundational model coming from Europe 👏🏻 Shall we add support in our app having in consideration it’s big size and the performance related to the already compatible Qwen3.6 35B A3B?
1
1
204
huggingface.co/Aleph-Alpha/k…
Aleph-Alpha/Kolibri-1 · Hugging Face

Aleph-Alpha/Kolibri-1 · Hugging Face

Text Generation model with 78B parameters. Download, explore files and run it on Hugging Face.

huggingface.co
14
52
6
818
145,652
We are REALLY proud to announce that MLXHub now allows you to run Qwen3.6 35B A3B on your iPhone 17 Pro (and other models). Big size is no longer an issue with MoE 🫡 Download the app for free: apps.apple.com/app/id6766485…
1
1
91
We are proud to announce that MLXHub is now officially supporting @IFM_AI K2-Horizon 7B and 36B A4B MoVA MLX conversions on your iOS and iPadOS devices! You can find them here: huggingface.co/collections/D…
K2 Horizon MLX Quants - a DreamFoundries Collection

K2 Horizon MLX Quants - a DreamFoundries Collection

MLX Q4, Q6, and Q8 conversions of IFM's K2 Horizon models, created for MLXHub with the DreamFoundries mlx-lm fork.

huggingface.co
1
1
1
165
BREAKING NEWS: Version 2.2.0 is finally approved and ready to use in the AppStore, go grab it! It comes with tons of new adjustments and features, like NAND Streaming, Tools & Skills, Memory, etc…
1
1
50
Whaaaat? No waaay! Who could have ever imagine that? And from @Meta ? Yeah, we guess there are in fact no surprises here, so if you care about your data and privacy, but wanna have your personal AI System, check MLXHub 🫦 Link in the profile 😏
Ray Wong (@raywongy) profile pictureRay Wong
@raywongySep 27
Deleted Muse after seeing this post on Threads about how it told some Facebook Marketplace sellers the guy’s address and they showed up at his door Dangerous and creepy This would have been 1000x worse if the person was a woman
Image from Ray Wong (@raywongy)
49
In a world where foreign governments can decide to cut or limit the availability of services such as AI (a competitive advantage today), having the ability to use local and private intelligence that you possess is not only recommended, but NECESSARY. apps.apple.com/es/app/mlxhub…
1
1
1
166
All 15 lifetime licenses had been delivered 🫡 Congratulations to the lucky users!
MLXHub (@MLXHub) profile pictureMLXHub@MLXHubJul 15
The first 15 followers will receive a special lifetime promo that will enable them to use MLXHub features free forever! Download the app: apps.apple.com/app/mlxhub-ia…
1
36
We are working hard to provide our users with SOTA local AI features. For that reason we cancelled version 2.1.0 of our app and moved on to 2.2.0, with huge chat improvements in it. Now in review by @Apple Version 2.3.0 in active development, we will focus in Voice 🫦
48
Did he said Hub? Local AI? 👀 #iPhoneDuo
55
UPDATE 🕦: The next version of MLX Hub, which includes support for Tools & Skills (among Web Search functionallities, Health, Music, Calendar, Reminders, etc) is being currently hold back by @Apple regarding non-existing ChatGPT integration within the app 🤷🏼‍♂️ we’ll keep you posted
1
1
2
109
more replies
That’s the version 2.1.0, rejected as of now 6 times by the same copy/pasted reasons. The version 2.2.0 with incredible changes is waiting in the queue to be uploaded to review, but we are working against the clock with the review team stuck somewhere.
1
15
We kindly apologize to our customers for the delays on providing suggested features and fixes, and the time passing without them to be able to enjoy the newly developed features. 5-star reviews on the AppStore helps us gain muscle, and your patience and support are appreciated!
9
Load more