Free tip of the day, if you have any kind of self improvement loop, show this paper https://t.co/F9y1RBJgW0 to the model and ask it to incorporate it.
Then when we meet in person one day, you can buy me a beer as a thank you.
🚨 JAPAN COULD CRASH MARKETS NEXT WEEK
Another US-Japan yen intervention could come as early as next week.
The last one (1998) was followed by a 20% crash in US stocks.
This is 50-100x bigger. 🫣
Meet Kimi-K3-0.40B, a compact yet powerful feature extraction model that's taking the ML community by storm. With 17K downloads, it's built on the Kimi-K3 base and fine-tuned for precision. Perfect for turning raw text into rich embeddings. Let's dive in! #AI#NLP
@SarvamAI
Lmao Cant even build your own harness from scratch.
https://t.co/RnHO4bdoUw
Building on top of an oss project and calling it indigenous coding agent lmao 🤡🤡🤡🤡
DeepSeek silently updated their changelog with a new V4-Flash upgrade 1 hour ago.
Their new Terminal-Bench score is 82.7, a massive +25.8 point leap from its initial April preview score of 56.9.
Currently only available via their API, open weights release will follow shortly.
DeepSeek-V4-Flash is updated.
«DSBench-FullStack is an internal full-stack development test set, and DSBench-Hard is an internal Coding Agent hard-problem test set»
«The official release of DeepSeek-V4-Pro will follow soon»
These scores are… quite a lot better than Preview.
@deepseek_ai team is seriously cracked. Damn.
They made improvements to their architecture and delivery of DeepSeek v4 flash, making it one of the cheapest and best models to use right now.
Almost 1/10th cost and so small compared to glm5.2 matching opus 4.8 crazy.
@VortexPM666@SarvamAI A simple attribution or changelog would have been great.
Also these days everyone can create their own harness except this lab which has more compute and gpu cards than any Chinese lab lol.
@Kimi_Moonshot ‘s team made kimi 3 with 5-8 year old Nvidia cards lmao.