Anthropic partners with U.S. National Labs for first 1,000 Scientist AI Jam

Source: Anthropic Date: 2025-02-28 URL: https://www.anthropic.com/news/anthropic-partners-with-u-s-national-labs-for-first-1-000-scientist-ai-jam

Summary

Anthropic participated in the DOE’s inaugural “1,000 Scientist AI Jam,” where scientists across US National Laboratories evaluated frontier AI models for scientific research and national security applications. Anthropic launched Claude 3.7 Sonnet at the event, describing it as “the first hybrid reasoning model on the market.” Builds on existing DOE/NNSA partnership, including classified environment testing in April 2024.

Implications

• Government/defense / AI for science thread. The 1,000 Scientist Jam is the DOE’s formal evaluation process for deciding which AI models get deployed at national labs. Anthropic launching Claude 3.7 Sonnet at the event is coordinated product timing — the model drops exactly when 1,000 government scientists are evaluating it.
• “First hybrid reasoning model.” This is the strongest capability claim in the Claude 3.7 Sonnet launch (Claude Code dropped with this model). “Hybrid reasoning” — switching between standard and extended thinking — was a meaningful differentiation from GPT-4 Turbo at the time.
• Classified environment testing (April 2024). The April 2024 classified lab testing is one of the few public disclosures of Anthropic operating in classified infrastructure. This establishes a track record for subsequent national security contracts.
• Watch: LLNL deployment scale (10,000 daily users cited in later signals); whether the Jam produces published evaluation results; how Claude 3.7 performs vs. other models in the scientific research use cases tested.