Claude Supercharges Biotech Models 4x

🔥 What happened Anthropic had Claude optimize over 30 open-source biomolecular models in under four weeks – averaging 4x speedup with minimal precision loss. A new low-memory mode enables prediction of systems over 10,000 tokens on a single NVIDIA GPU. The code is open-sourced, alongside a protein design competition with up to $1M in Claude credits. 💡 Why it matters Previously, one protein design target cost up to $10,000 (~2,500 H100 hours). Now, similar results are possible with two orders of magnitude fewer GPU hours – democratizing drug discovery. The FlashPairformer triangle-attention kernels beat the industry standard by up to 2.9x. ⚡ Our take Anthropic proves AI isn't just training models – it's overhauling the entire scientific toolchain. If you're not paying attention, you'll fall behind in AI-driven biotech.