←
Qualcomm Developer2,249 posts
Qualcomm Developer avatar

Qualcomm Developer

@Qualcomm_Dev

Build what’s next. We’ll bring the stack. qualcomm.com/developer

◷ Joined December 2011▧ 337 Photos and videos
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·54m

Did you know you can run native PyTorch with vLLM on @Qualcomm Dragonfly? Watch Qwen3-VL-32B, running locally in full FP16 via the open-source vLLM-QAIC plugin. #PyTorch #vLLM

1:06
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 17

What happens if a small decision model gets first look at every AI request? Lucas da Silva tried it with open-source Laya, running the full model on the Snapdragon X Elite NPU. It can spot private data, triage requests and decide whether work stays local or goes to the cloud. No quantization. No retraining. Just the model running on the NPU at 8.6 ms per decision, vs. 36 ms on CPU, with 20x+ less energy per decision. A promising pattern for fast, local decision-making before bigger models enter the workflow.

1:03
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 16

Are you joining us at Hexagon Day 2.0? With live demos, technical sessions, and ample networking possibilities, you’ll learn what’s new, what’s next, and what’s possible with Hexagon. This is your last chance to reserve a spot! Numbers are limited: bit.ly/4zER3kd

Post attachment
Snapdragon avatar
⟳ Qualcomm Developer retweeted
Snapdragon@Snapdragon·Nov 12

These are the coolest (they're powered by Snapdragon AR1 Gen 1 😎)

Introducing intelligent eyewear, built in partnership with Samsung and our iconic eyewear brands, Gentle Monster and Warby Parker. Now available for pre-order. Learn more at goo.gle/glasses_x

0:29
Modular avatar
⟳ Qualcomm Developer retweeted
Modular@Modular·Nov 12

Chris Lattner’s #PyTorchCon keynote is now on YouTube. @clattner_llvm covers running models across more hardware with Mojo and MAX, and building an open developer ecosystem with @Qualcomm_Dev and @PyTorch. youtube.com/watch?v=2XXf_iL1…

Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 11

Somewhere between the frame you submitted and the GPU crash is the thing you actually need to find. Adreno Crash Insight captures GPU crash data to help connect those dots: failure → frame → workload. bit.ly/4jbjUH5

Post attachment
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 10

Want to experiment with LLM-driven NPCs in games? 🕹️ Our gaming team just introduced MindSand, an open-source project developers can use to explore natural-language commands, NPC personalities and runtime conversations. See how the team put it to work in Starship Troopers: Extermination: bit.ly/3WnyJxw

0:47
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 6

From shipping edge AI to building devices that understand their surroundings - that’s where our developer conversations are going at #WebSummit on November 10: 11:00 WET: Manny Singh on prototype → production with Qualcomm Dragonwing 16:00 WET: Ziad Asghar on physical AI, perception + context In Lisbon? Join us. Elsewhere? Follow along.

Post attachment
Modular avatar
⟳ Qualcomm Developer retweeted
Modular@Modular·Nov 4

LFM3 from @liquidai is now running on MAX. 🧑‍🚀 All 6 models in the suite are supported, across text, audio, vision, and omni. The implementations are open source, so you can read the code, run it, and build on it: max.modular.com/api/python/p… Which modality are you trying first: text, audio, vision, or omni?

Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵

2:17
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 3

We put Qualcomm Dragonwing IQ-9075 EVK in @nikodembartnik’s hands. The result? An edge AI experiment that gives home security a little more bite. 👀📢

Standard security cameras are useless - they just record people stealing your stuff. So I teamed up with @Qualcomm_Dev to build an AI camera system that screams personalized warnings at intruders instead. 🔈👀

1:05
Qualcomm Developer avatar
Qualcomm Developer@Qualcomm_Dev·Nov 2

“Package thief in the blue shoes—get out of here!” 🗣️ That’s just one of the infinite things this edge #AI doorbell camera can say to would-be porch thieves. Watch as developer @nikodembartnik build this home security system using @Qualcomm Dragonwing IQ-9075 EVK: bit.ly/4yS8jBz

0:31