Grok 4.6 just dropped and will be on LiveBench shortly
- same price as Grok 4.5 which is great
Not sure how much it really improves on 4.5. We will see
New model releases are going to pick up again
Models that I am most excited about
- Grok 4.6
- GLM 5.5
- Deepseek v4 Pro
Pricing will be key for adoption
People who are pushing to "Pause AI" are utterly clueless
In essence, they are advocating for a permanent underclass
"Pausing AI" simply translates to a small set of folks in frontier labs having access to large 10T+ models
Everyone else gets nothing and will eventually be automated away by super intelligence
Basically, the "pause AI" types are simply advocating for monopolies that will control the rest of human kind 🤷
Yikes - Anthropic is going to add an invisible watermark to all the text they generate
The good news is Claude was never that great at text generation
You should simply avoid using them for things like report writing
SOL and Grok are much better
Large companies including Microsoft and Amazon are stepping back from developing frontier models
The latest Google rumor is 3.5 Pro is canceled as well
Gemini 4.0 is easily months away
Big mistake to exit the AGI and large model race
🚨 A NEW OPEN-SOURCE DRAGON MODEL DROPS TONITE
It is based on Kimi K3 and improves on agentic capabilities!!
A 2.8T model for coding and agentic tools. Watch out for it on Hugging Face
Meta is sooo back and is now on the top of the US open-source leaderboard 🚀
Muse-Spark 1.2 is a very good small model that is BETTER than DeepSeek Flash
If Meta goes back fully to their open source roots, I expect them to beat Kimi K3 and GLM 5.5 in 3 months
The future will be dominated by two types of models
ULTRA SMART - large extremely intelligent 10T models that can do very complex work
ULTRA CHEAP - near free intelligence that can handle pretty much everything else
Today’s equivalents are Fable and Luna / Deepseek Flash
Lots of rumors about OpenAI's new model Astra and Fable 5.1
- great at long running agentic loops
- Astra is better than Fable 5
- Anthropic to release a Fable 5.1 to compete with it
- both models in safety testing
- will likely release after GLM 5.5
TBH, you just have to spend 10x more time making an open-source model work in production
- K3 is super slow
- Deepseek flash will spin on more complex problems
- GLM and Flash don't have visual capabilities
I am close to totally giving up and going back to closed frontier models 😓