So big labs have locked in so much compute that prices are now 3x'ing in months on RAM / VRAM / etc - they are going to price themselves out of scaling?
We are also now working closely with cline to make Hermes even stronger in Cline with direct work on training and optimizing built around their coding harness!
"Follow the gradient" That's great.
We are learning so much from open source, shipping, screwing up and figuring it out.
Plans are good, reality is better
A big milestone for Hermes.
We did a lot of work to make a frontier level openmodel that does not dictate what expression you can elicit from the model.
Super strong at math, coding, STEM, and creativity.
Model Weights: https://t.co/ft01mebGW4
Check it out 👇
Anecdotally, I’ve found the people most vocal and showy about grinding hard (9-9-6) tend to have less throughput than a garden variety workaholic.
I suspect this is because they’ve internalized endurance pace all the time. And loose the ability to sprint when needed.
Crazy week!
@moritzthuening ported his project from Wormhole to Blackhole with 1-line of code change.
@OnDemandai has a killer agentic workflow available in the cloud and on-prem for @tenstorrent users
@gokoyeb continues to expand TT offerings
🥹 it’s a beautiful thing
Another image -> 3D -> v2v (wan 2.1) render. The quality you can get is awesome. And because it's 3D mid workflow you can do arbitrary camera position / movement etc.
Still 10 hour rendering time 🤡
Congrats to our post training team who worked on the Hermes 3's dataset - @teknium, @nullvaluetensor, and outside contributor @intrstllrninja - on creating the now #1 Trending dataset on HuggingFace!