Interested in AI, Machine Learning and all things Tech. Tweets are my own. RT is not endoresment.
Joined June 2020
- Tweets85
- Following67
- Followers26
- Likes186
It's been a lot of fun working with the @TencentHunyuan team to push the boundaries of low precision compute. The models were expertly crafted for high quality and demonstrate exceptional performance on CPUs with SME2.
@Arm Accelerating translation for Tencent users 🚀
🚀 Tencent Hy Translation just landed.Powered by Hy-MT2.
33 languages. 5 Chinese minority languages & dialects.
Voice. Photo. Full offline — on-device, no network required.
Already live in 12 countries and regions.
Travel, drive, work or read abroad. Accurate. Natural. Always available.
⏬ ⏬⏬ apps.apple.com/us/app/tencen…
Ronan Naughton retweeted
⚡ Meet Qwen-Audio-3.1! ASR, TTS & Realtime are fully upgraded, joined by two new models: TTS-Next for audio creation and ASR-Next for audio understanding.
Five models, one complete audio stack: understanding, generation, interaction & creation.
Plus big price cuts across the lineup: TTS ~70% off, Realtime ~85% off, and ASR up to 95% off.
Highlights: 🥳
- ASR: stronger multilingual & dialect recognition, plus native polishing that auto-removes fillers & repetitions for cleaner, more logical transcripts.
- ASR-Next: supports multi-speaker ASR with speaker labels, timestamps & aligned transcripts, and understands emotions, ambient & machine sounds for sound captioning, event localization, audio QA & reasoning.
- TTS: multilingual & dialect synthesis with natural cross-lingual voice transfer; control emotion, speed & style via simple instructions.
- TTS-Next: unified LM + diffusion framework generating voice, sound effects & background audio in one pass for audiobooks, podcasts, games & ads.
- Realtime: speak & listen at once with anytime interruption, just like a real call; it even slows down and responds empathetically when it senses a low mood.
Unlock the full potential of Qwen-Audio-3.1! 👇
- Blog: fun-resource-shanghai.oss-cn…
- Qwen-Audio-3.1-ASR:
qwencloud.com/models/qwen-au…
- Qwen-Audio-3.1-Realtime:
qwencloud.com/models/qwen-au…
- More APIs: coming soon @qwen_cloud
Ronan Naughton retweeted
A defining moment for AI compute.
See how collaboration across hardware, software, and systems is enabling better computing platforms for everyone.
🎥 Join the livestream | March 24 okt.to/o9ArRc
#ArmEverywhere
Developers: Welcome to a new era of portable performant on-device AI with Arm Lumex and SME2.
• 5x Performance 🚀
• 3x Efficiency🔋
SME2 Unlocks edge AI use cases such as Translation, Speech Recognition, Screen Description, SLMs and many more.
@Alipay #Vivo #KleidiAI
🆕 Meet Arm Lumex, our most advanced CSS platform for flagship smartphones and next-gen PCs.
Uniting SME2-enabled CPUs with new GPUs and system IP, it delivers up to 5x faster AI performance, fueling smarter, quicker experiences on billions of devices: okt.to/Uc3koJ
📣 @Arm and @Microsoft are delighted to announce the integration of #KleidAI into @onnxruntime. This means more efficient AI inference, faster interactions, and better performance for AI workloads across Windows and Android applications.
Check out the blog: lnkd.in/eyK6myM5
There is a lot more to come... 🚀
Congratulations to Alibaba on the release of #Qwen3 model which demonstrates SOTA AI performance. Arm and Alibaba have been collaborating closely to accelerate Qwen3 on CPU via Arm KleidiAI, providing exceptional AI performance to Alibaba users.
Find out more ⬇️:
Ronan Naughton retweeted
🚘 Arm Kleidi has arrived in automotive!
Designed to deliver performance optimizations for AI inference workloads running on Arm CPUs, it's driving faster response times & accelerating time-to-market across new applications for software-defined vehicles: okt.to/vbrP9V
Arm #KleidiAI and Alibaba MNN have teamed up to accelerate multimodal workloads on Arm-based phones.
This unlocks enhanced multimodal AI experiences and faster response times - as shown at #MWC25:
newsroom.arm.com/blog/arm-kl…
Text-to-audio AI is now possible entirely on smartphones on the CPU thanks to @Arm and @Stability AI. Type a description, and generate the perfect audio in seconds, with yet more performance optimizations planned for the future. Get the details: newsroom.arm.com/blog/stabil…
#KleidiCV takes its place in mobile with #OpenCV 4.11 integration. 🤝
Developers... check out the speedups 🚀
We’ve worked with OpenCV on the first ever KleidiCV integration! 🎉
With 4x performance uplifts on OpenCV 4.11, in Android KleidiCV is enabling faster responses across key computer vision tasks on mobile with no extra developer effort.
Get the details: okt.to/ASLwkc
Ronan Naughton retweeted
I think running llms on the CPU is actually a really good way to look at year-over-year performance gains on a CPU
So here is @Snapdragon 8 gen 2, 8 gen 3, and 8 Elite running R1 Distill Qwen 1.5B in the PocketPal AI benchmark
8G2: 14.18 t/s
8G3: 24.62 t/s
8 Elite: 33.58 t/s
Ronan Naughton retweeted
Jensen just unveiled the new NVIDIA Jetson Orin Nano Super, optimized for on-device AI processing in robots,
boasting 70 trillion operations (INT8) per second, 102GB/s memory bandwidth, 25 watts power consumption, and a price of $249.
A nice video from @garyexplains explaining the benefits of AI on CPU and showcasing LLAMA 3.2 1B running on a mobile device 📱.
Check it out ⬇️
youtube.com/@GaryExplains
📰 AI News of the day!
➡️ @AIatMeta release quantized LLAMA 3.2 models
➡️ Accelerated by Arm's #KleidiAI library
🤔 Find out more: community.arm.com/arm-commun…
A big thank you to the entire Executorch team at AI at Meta for this great collaboration. 🙏
LLAMA3.2 1B provides exceptional performance #onArm
Multimodal models will unlock new use cases for mobile users.
#Executorch and #KleidiAi will bring performance of these models to the next level. 🚀
Find out more: newsroom.arm.com/news/ai-inf…
Today is the day ➡ #PyTorchConf!
I will be there with @Arm.
Please reach out to me if you'd like to connect and hear about some of the GenAI solutions we are developing #onArm.
hubs.la/Q02LCtFZ0
@PyTorch
