𝗗𝗼 𝘆𝗼𝘂 𝗿𝗲𝗻𝘁 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲, 𝗼𝗿 𝗱𝗼 𝘆𝗼𝘂 𝗼𝘄𝗻 𝗶𝘁?
In the past days, we could access #Fable5, again and thought: #FrontierAI is starting to look less like a product you access, and more like a capability you may not be allowed to use.
Recent signals from the US suggest that access to the 𝗺𝗼𝘀𝘁 𝗮𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗺𝗼𝗱𝗲𝗹𝘀 𝗰𝗼𝘂𝗹𝗱 𝗯𝗲 𝗿𝗲𝘀𝘁𝗿𝗶𝗰𝘁𝗲𝗱, 𝗽𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹𝗹𝘆 𝗲𝘃𝗲𝗻 𝘁𝗼 𝘁𝗵𝗲𝗶𝗿 𝗼𝘄𝗻 𝗰𝗿𝗲𝗮𝘁𝗼𝗿𝘀. If that trend continues, any model reaching “𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗹𝗲𝘃𝗲𝗹” capability may face similar controls, possibly tied to identity or nationality verification.
But there is a parallel story unfolding, one that matters just as much for builders, enterprises, and policymakers: 𝗢𝗽𝗲𝗻-𝘄𝗲𝗶𝗴𝗵𝘁 𝗺𝗼𝗱𝗲𝗹𝘀 𝗮𝗿𝗲 𝗰𝗮𝘁𝗰𝗵𝗶𝗻𝗴 𝘂𝗽. 𝗙𝗮𝘀𝘁.
A new system, 𝗚𝗟𝗠 5.2 is a 744‑billion‑parameter mixture‑of‑experts model created by 𝗭.𝗮𝗶 (formerly Zhipu AI), a Chinese AI lab.
It shows how quickly the gap is closing:
- A significant leap in under 3 months
- Strong performance across coding, math, and general reasoning
- Noticeably ahead of most open models, approaching frontier level in some benchmarks
𝗜𝘁 𝗶𝘀 𝗻𝗼𝘁 𝗳𝘂𝗹𝗹𝘆 𝗼𝗻 𝗽𝗮𝗿 𝘄𝗶𝘁𝗵 𝘁𝗼𝗽 𝗽𝗿𝗼𝗽𝗿𝗶𝗲𝘁𝗮𝗿𝘆 𝘀𝘆𝘀𝘁𝗲𝗺𝘀, 𝗯𝘂𝘁 𝗶𝘁 𝗶𝘀 𝗰𝗹𝗼𝘀𝗲 𝗲𝗻𝗼𝘂𝗴𝗵 𝘁𝗼 𝗰𝗵𝗮𝗻𝗴𝗲 𝘁𝗵𝗲 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻.
There are trade-offs:
- 𝗩𝗲𝗿𝘆 𝗹𝗮𝗿𝗴𝗲, with hundreds of billions of parameters
- 𝗘𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 to run without serious infrastructure
- 𝗛𝗶𝗴𝗵𝗲𝗿 𝘁𝗼𝗸𝗲𝗻 𝘂𝘀𝗮𝗴𝗲 in some cases
But the trajectory is clear, and the implications are strategic. We may be heading toward a world where:
- 𝗙𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗔𝗜 𝗶𝘀 𝗴𝗮𝘁𝗲𝗱, 𝗿𝗲𝗴𝘂𝗹𝗮𝘁𝗲𝗱, 𝗮𝗻𝗱 𝗰𝗲𝗻𝘁𝗿𝗮𝗹𝗶𝘇𝗲𝗱
- 𝗢𝗽𝗲𝗻-𝘄𝗲𝗶𝗴𝗵𝘁 𝗔𝗜 𝗶𝘀 𝘀𝗹𝗶𝗴𝗵𝘁𝗹𝘆 𝗯𝗲𝗵𝗶𝗻𝗱, 𝗯𝘂𝘁 𝗮𝗰𝗰𝗲𝘀𝘀𝗶𝗯𝗹𝗲, 𝗮𝗱𝗮𝗽𝘁𝗮𝗯𝗹𝗲, 𝗮𝗻𝗱 𝘀𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻
That raises a fundamental question for organisations: Do you rent intelligence, or do you own it?
What we know from the #crypto world, “Not your keys, not your coins” becomes “𝗡𝗼𝘁 𝘆𝗼𝘂𝗿 𝘄𝗲𝗶𝗴𝗵𝘁𝘀, 𝗻𝗼𝘁 𝘆𝗼𝘂𝗿 𝗺𝗼𝗱𝗲𝗹”. It is no longer a fringe view. It is becoming a #riskmanagement principle.
If current progress holds, we could see near frontier-level open models within the next year. That would shift the balance of power away from a handful of providers, and toward anyone capable of deploying and operating their own AI stack.
For those working at the intersection of #AI, #cyber, and #geopolitical risks, this is one to watch closely.
Watch this very interesting video: https://www.youtube.com/watch?v=qks6dGQFd_c
Check out the following video for more information:
