FWIW – we work quite closely with all major Chinese labs, we don't distinguish between any. Wherever there's open source/ science, HF is there.
@reach_vb
-
DeepSeek’s 5.5M USD Breakthrough: Frontier AI Cost Revolution
By
–
Agreed. On a similar note, what are your thoughts on the recent DeepSeek release? The frontier *just* costs 5.5M USD (+ a cracked team) – and I'm here for it!
-
DeepSeek funding strategy to acquire GPUs and top talent
By
–
Maybe I'm biased, I want DeepSeek to make metric fk ton of money – as much as possible, be it through API fees or some other way – I want them to have enough money to acquire GPUs/ talent, whatever it takes to get to the next frontier as fast as possible
-
FP8 Precision Format Discussion in Machine Learning
By
–
My understanding from the paper is that it's fp8. Maybe @zizhpan can confirm
-
Text Generation Inference v3.0.1 release fixes reported issue
By
–
this should not happen, can you try with the latest tagged image – 3.0.1 happy to flag it to the team if it still doesn't work! sorry for the inconvenience! https://
github.com/huggingface/te
xt-generation-inference/releases/tag/v3.0.1
… -
Frontier Model Development Cost Drops to 5.5 Million USD
By
–
Scarcity breeds Innovation – cost to build a frontier model – 5.5 Million USD In a way, it's the maximum it'd be (Note: H800s have ~2x slower chip-to-chip data transfer) This cost, will only go down further and further as we continue to find newer walls to scale!
-
Qwen and Meta GPU Mobilization in AI Competition
By
–
I wouldn't discount Qwen or Meta either – at least the latter is mobilising metric fk ton of GPUs
-
Open Science Momentum Builds Into 2025
By
–
4 days to the end of 2024, open science is winning! Bring it on in 2025!
-
TGI Native Support Available Through Version 2.5
By
–
We do support it natively in TGI (till v2.5) – top of the list
-
DeepSeek API Direct Consumption Performance Evaluation
By
–
That said, what’s wrong w/ consuming directly from DeepSeek API, it looks pretty fkn fast to me.