ResearchSep 16, 2026 · 02:006 sources
Periodic Labs 1 Trillion Parameter Neon Model Beats GPT-6 Astra in Science
From huggingnews.com
Periodic Labs developed Neon, a 1 trillion parameter open-source LLM trained with 1,300 H200 GPUs and months of wet lab data, which outperformed GPT-6 Astra and Fable on benchmarks for superconductors and magnets. The model uses mid-training and reinforcement learning and achieved 4.1x training throughput over a Megatron baseline and 2.5x faster decoding with cluster utilization above 95%. Periodic Labs contributes its technical improvements to open-source projects including Megatron-LM, SGLang and Miles.
Read the sourceSign in to suggest
Sources
+ 4 more
- SUPPORT@liamfedus“train specialized scientific models with strong performance on relatively modest compute compared to frontier systems”x.com
- SUPPORT@liamfedus“4.1× training throughput vs. our Megatron baseline 2.5× faster decoding 95%+ cluster utilization”x.com
- SUPPORT@khoomeik“We midtrain + RL’d a trillion param LLM to analyze experimental data from our superconductor lab”x.com
- SUPPORT@_jasonwei“wet lab data enables a specialized model to beat gpt-6 astra at a task at the frontier of science!”x.com
#periodic-labs#neon#llm#open-source#benchmark#chips#gpt-6-astra#science-benchmarkMarkdown
Content history (2)
- 2026-10-04 · Summary · vi · Replaced stored textPeriodic Labs công bố mô hình Neon với 1 nghìn tỷ tham số, đạt kết quả vượt trội so với G…Một cơ sở có thông lượng cao ở Menlo Park đã phát triển trí tuệ nhân tạo chuyên biệt, vượ…
- 2026-10-04 · Title · vi · Replaced stored textMô hình Neon 1 nghìn tỷ tham số của Periodic Labs vượt trội GPT-6 Astra trong lĩnh vực kh…Neon model 1 nghìn tỷ tham số của Periodic Labs vượt qua GPT-6 Astra trong khoa học
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today