--- format: "aidr-story-markdown/v1" id: "f7ff4ff9e1217abd73e788a64946681fe94e3e04ad1667e87c8be04d0d960010" canonical_url: "https://aidr.today/f7ff4ff9?lang=en" title: "Periodic Labs 1 Trillion Parameter Neon Model Beats GPT-6 Astra in Science" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-09-15T19:00:43.000Z" category: "Research" topics: ["periodic-labs","neon","llm","open-source","benchmark","chips","gpt-6-astra","science-benchmark"] source_urls: ["https://huggingnews.com/ai/periodic-labs-1-trillion-parameter-neon-model-beats-gpt-6-astra-in-scien-91f55089","https://x.com/LiamFedus/status/2099896055030501702","https://x.com/LiamFedus/status/2099896057299656950","https://x.com/LiamFedus/status/2099896061548487152","https://x.com/khoomeik/status/2099898669915132274","https://x.com/_jasonwei/status/2099907698708095210","https://marketbrief.now/ai/periodic-labs-1-trillion-parameter-neon-model-beats-gpt-6-astra-in-scien-91f55089"] summary: "A high-throughput facility in Menlo Park developed a specialized artificial intelligence that outperforms general frontier systems in the analysis of semiconductor materials. Periodic Labs calls the tool Neon, a 1 trillion parameter open-source LLM trained via mid-training and reinforcement learning. The model used 1,300 H200 GPUs and months of wet lab data to surpass GPT-6 Astra and Fable on benchmarks for superconductors and magnets. The laboratory infrastructure reached 4.1x training throughput over a Megatron baseline and 2.5x faster decoding with cluster utilization above 95%. This architecture creates a cycle where AI manages lab operations and learns from the resulting data to inform future experiments. Periodic Labs contributes its technical improvements to open-source projects including Megatron-LM, SGLang and Miles." --- # Periodic Labs 1 Trillion Parameter Neon Model Beats GPT\-6 Astra in Science > [Open the canonical story]() **Published:** 2026-09-15T19:00:43.000Z **Category:** Research **Topics:** periodic\-labs, neon, llm, open\-source, benchmark, chips, gpt\-6\-astra, science\-benchmark ## Summary A high\-throughput facility in Menlo Park developed a specialized artificial intelligence that outperforms general frontier systems in the analysis of semiconductor materials\. Periodic Labs calls the tool Neon, a 1 trillion parameter open\-source LLM trained via mid\-training and reinforcement learning\. The model used 1,300 H200 GPUs and months of wet lab data to surpass GPT\-6 Astra and Fable on benchmarks for superconductors and magnets\. The laboratory infrastructure reached 4\.1x training throughput over a Megatron baseline and 2\.5x faster decoding with cluster utilization above 95%\. This architecture creates a cycle where AI manages lab operations and learns from the resulting data to inform future experiments\. Periodic Labs contributes its technical improvements to open\-source projects including Megatron\-LM, SGLang and Miles\. ## Sources - [Story source]() - [Story source]() - [Supporting source]() - [Supporting source]() - [Supporting source]() - [Supporting source]() - [Story source]()