#AI-85X
A vision #AI model #VLM that is 85x faster, 3x sm, 4x less tokens, sounds really awesome.

Until one realizes it's only available by overpaying for a MFG limited brand new hardware that's 'walled garden ecosys', closed source, mfg tied & by a Trump d°°k riding co.
FU #Apple
youtu.be/LY9gYYuHYsU?...
Apple’s New AI SHOCKS The Industry With 85X More Speed (Beating Everyone)
YouTube video by AI Revolution
youtu.be
September 3, 2025 at 11:12 PM
acordei umas 85x durante a noite ai já jurava que hoje era quarta disgraça
April 15, 2025 at 11:50 AM
#Pinterest expanded its #Nvidia partnership, now built on Blackwell GPUs, to power AI discovery. Pinterest Assistant runs 25x more visual context per request, with ~85x faster response startup and 7.3x faster latency. #MarketingSky #AI
Pinterest announces new deal with Nvidia
The move builds on an existing relationship between the two companies and seeks to improve the app’s AI-powered discovery process.
www.socialmediatoday.com
September 21, 2026 at 10:59 PM
Marvell Technology Faces Valuation Concerns Despite Strong AI Growth: Marvell Technology stock fell roughly 25% from recent highs despite management projecting revenue growth acceleration from 40% to 45% and making $1 billion in prepayments to secure manufacturing… #MarketUpdates Follow Us for more
Marvell Technology Faces Valuation Concerns Despite Strong AI Growth
Marvell Technology stock fell roughly 25% from recent highs despite management projecting revenue growth acceleration from 40% to 45% and making $1 billion in prepayments to secure manufacturing capacity. The pullback reflects valuation concerns as the stock trades at 85x trailing P/E, well above peers, even as the company expands its AI infrastructure business through deeper NVIDIA partnership and custom chip design for hyperscalers.
quantli.com
July 9, 2026 at 9:50 PM
My AI agent memory layer was taking 340ms.
Rewrite the hot path in Rust: 4ms.
85x faster. Saturday afternoon. 47 lines of code.
The bottleneck was never the LLM API. It was my own Python.

tinyurl.com/39mvxmd8
47 Lines of Rust. 85x Faster Agent Memory
I stopped blaming the LLM API. The problem was my own code.
tinyurl.com
May 27, 2026 at 11:31 AM
Apple Released FastVLM: A Novel Hybrid Vision Encoder which is 85x Faster and 3.4x Smaller than Comparable Sized Vision Language Models (VLMs) Introduction Vision Language Models (VLMs) allow both ...

#AI #Paper #Summary #Artificial #Intelligence #Computer […]

[Original post on marktechpost.com]
September 2, 2025 at 6:13 PM