Periodic labs beat gpt-6 astra on its own hard science benchmark with 1,300 h200s. the compute arms…
periodic labs beat gpt-6 astra on its own hard science benchmark with 1,300 h200s. the compute arms race just found a cheaper path.
the next frontier lab is a garage with a credit line.
Context
Periodic Labs' post of 15 September 2026 says its model Periodic Neon outperformed GPT-6 Astra and Claude Fable 5.1 at lower cost on FrontierXRD, a Periodic internal evaluation of 134 samples from its labs, with 55.3% success for Neon, scored by an LLM-judge ensemble, all models run in Periodic's own harness. It says the final training run used 1,300 H200 GPUs, against Astra's reported 100,000 or more Blackwell GPUs, a count it cites to a report that was not inspected.
This is a vendor-reported result on Periodic's own benchmark. On its own hard science benchmark refers to Periodic's FrontierXRD, not OpenAI's; the premise that it is Astra's own benchmark has no support, and OpenAI's Astra page text read contains other science evals but no FrontierXRD. The Astra GPU count is unverified. The next frontier lab being a garage with a credit line is the author's opinion.
Related work
- AI infrastructure at Periodic ↗Snippet only.
- GPT-6 Astra (OpenAI) ↗Partly read.
Watch next
- Independent replication or release of FrontierXRD.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 18 September 2026 at 04:05 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →