● ModelsNov 1, 2026 · 22:373 sources

Carbon-X 14B Model Matches GPT-6.1 Sol in DNA Tasks at Lower Cost

From marketbrief.now

AI;DR

Carbon-X is an open-weights AI tool that performs biological sequence analysis and variant effect prediction using natural language. The 14B parameter model performs within one point of the proprietary GPT-6.1 Sol on DNA benchmarks at more than 1,000 times lower inference costs. Carbon-X was trained on more than 300B tokens of genomic data and 300k supervised and reinforcement learning samples.

Read originalSign in to suggest

Sources

+ 1 more
  1. SUPPORT@_lewtun“Mid-training: 300B+ tokens of genomic sequences, papers and web pages mixing DNA and text.”x.com
#llmMarkdownScore 1.13.7k tokens
Image: marketbrief.now
Read original