People sleeping on DeepSeek Math V2 are completely missing the shift in how models actually learn. Achieving gold medal status at the IMO while being open source? That is massive. The "verifier" approach isn't just a technical detail; it is the difference between guessing and reasoning. If you think proprietary models are the only ones capable of complex logic, you aren't paying attention. This is pure utility available to everyone. MODEL: https://huggingface.co/deepseek-ai/DeepSeek-Math-V2 PAPER:
- 0 replies
- 0 recasts
- 0 reactions
I just completed "OP Airdrop #5" on Layer3 - https://app.layer3.xyz/quests/op-airdrop-5
- 0 replies
- 0 recasts
- 0 reactions
Did a Tyson Pushup. It’s amazing how some people can do it like 50-100 reps. https://youtu.be/j-oSVFU2eMc?si=jioaL2sse73AdX3E
- 0 replies
- 0 recasts
- 0 reactions