osworld-verified bench tests ai on real computer tasks – opening apps, filling forms, navigating browsers. each task is pass/fail. the % is simply how many tasks the model completed successfully
Nick Trenkler
1 min read
the information just dropped a piece claiming that, and a lot more the technique copies capabilities from a trillion-parameter model down into something light enough to run on-device.
Addy Crezee
1 min read
bytedance is now building its own custom cpus to power its ai infrastructure – and they're not alone. google, amazon, and microsoft are all doing the same thing. the
Nick Trenkler
1 min read
microsoft has released its first strong ai model – mai-image-2.5. here's the breakdown mai-image-2.5 hit #3 on the text-to-image arena with
Nick Trenkler
1 min read
benchmarks (mimo vs deepseek): • gdpval-aa (general agent elo): 1581 vs 1554 • τ³-bench (tool-use): 72.9 vs 71.8 • claweval (function calling): 63.8 vs 59.8 • humanity&
Addy Crezee
1 min read
our ai host mira just pulled together the key details on openrouter's series b round • the token economy keeps scaling. openrouter now processes 25 trillion tokens per week,
Nick Trenkler
1 min read
thehype analyzed a post by @MiniMax_AI's head of engineering announcing the m3 model and its architecture. here's what we've found out in most
Nick Trenkler
2 min read
hermes is now officially faster than codex. here's how they did it hermes was losing to codex 6–5 across 11 benchmark tasks measuring real-world cli speed
Nick Trenkler
1 min read
built by @mvanhorn and published through @ppressdev tesla model y is the top-selling ev in california, florida, and texas. roughly 103k vehicles were sold in the us in q1
Addy Crezee
1 min read
- on april 18th, elon said the next model (grok 4.4) would have 1t parameters and release in early may - grok 4.5 was planned to follow late may at
Nick Trenkler
1 min read
another bunch of unsolved math problems was solved by ai. here's how google's agent solved 53+ open math problems. what it solved: • 9 erdos problems out
Addy Crezee
1 min read
openhuman gains 78% in stars, reaching 24,354 total stars agentmemory fits in this trend as well, growing 51% in stars. looks like persistent memory is here to stay here
Nick Trenkler
1 min read
deepseek has announced a steep price cut for its v4-pro model api. once its 75%-off promo ends on may 31, 2026, v4-pro settles at a quarter of
Nick Trenkler
1 min read
• 228k employees • 80k r&d (all in) • 25% of the other 148k = 37k ≈117k total claude users among microsoft employees • opus 4.7 pricing: $5/m input, $25/m
Nick Trenkler
1 min read
telegram just automated your dms. bots can now connect to user accounts and handle incoming dms on their behalf capabilities: • read messages in user-approved chats • reply as the user
Nick Trenkler
1 min read