This week has been marked by significant advancements in AI technology and notable shifts in leadership. Here's a breakdown of the key developments:
What Happened
Liquid AI released LFM2.5-230M, its smallest model yet, which runs on-device at 213 tok/s on a Galaxy S25 Ultra and 42 on a Raspberry Pi 5. Built on the LFM2 architecture, it targets tool use and data extraction, beating larger models like Qwen3.5-0.8B and Gemma 3 1B on instruction following.
DeepSeek open-sourced DSpark, a speculative decoding framework that attaches a draft module to existing DeepSeek-V4 weights. It pairs a parallel draft backbone with a lightweight Markov head to cut suffix decay, then adds confidence-scheduled verification that tailors how many tokens get checked to real-time GPU load.
Meta released Astryx, an open-source React design system built on StyleX. It pairs a CSS-variable theme cascade with a CLI and MCP server, so both engineers and AI agents build using the same API.
Meanwhile, Paul Meade, the Apple vice president in charge of the Vision Pro headset, is reportedly leaving the company to join OpenAI’s hardware team.
Why It Matters
These developments demonstrate the rapid progress being made in AI research and development. Liquid AI's LFM2.5-230M model showcases the potential for smaller, more efficient models that can run on-device, while DeepSeek's DSpark framework highlights the importance of speculative decoding in accelerating per-user generation.
Meta's Astryx design system, on the other hand, represents a significant step towards creating more accessible and user-friendly AI-powered tools. By providing a common API for both engineers and AI agents, Astryx has the potential to democratize AI development and make it more inclusive.
Key Numbers
- 230M: The number of parameters in Liquid AI's LFM2.5-230M model
- 213 tok/s: The speed at which LFM2.5-230M runs on a Galaxy S25 Ultra
- 42: The speed at which LFM2.5-230M runs on a Raspberry Pi 5
- 60-85%: The acceleration of DeepSeek-V4 per-user generation over MTP-1 using DSpark
Key Facts
- Who: Liquid AI, DeepSeek, Meta, OpenAI, Apple
- What: Released new AI models and frameworks, introduced open-source design system, announced personnel changes
- When: This week
- Where: Global
- Impact: Significant advancements in AI research and development, potential for more accessible and user-friendly AI-powered tools
What Comes Next
As AI technology continues to advance at a rapid pace, we can expect to see more innovative applications and use cases emerge. With the release of LFM2.5-230M and DSpark, we may see a shift towards more efficient and effective AI models that can run on-device. Meanwhile, Meta's Astryx design system has the potential to democratize AI development and make it more inclusive.
The departure of Paul Meade from Apple to OpenAI may also signal a new era of collaboration and innovation between the two companies. As the AI landscape continues to evolve, one thing is clear: the future of AI is bright, and we can expect to see many more exciting developments in the coming weeks and months.