The announcements today reveal three distinct competitive postures: a race to embed AI into enterprise workflows at scale, a scramble for infrastructure dominance as training and inference costs become the primary battleground, and a pivot toward public-facing safety and research positioning that mirrors regulatory and reputational concerns. OpenAI is showcasing cost efficiency and workflow automation across Oracle, LegalOn, and creative tools, the message is that their models are already cutting real spending and accelerating work, which matters more than abstract capability claims. Meanwhile, NVIDIA and AMD are locked in a technical specification war over which accelerators train and serve models faster and cheaper, with AMD publishing detailed performance breakdowns and NVIDIA committing $1 billion to position itself as the infrastructure partner for "super intelligence research," a framing that sidesteps the commodity chip competition by attaching hardware to national interest. Anthropic, by contrast, is broadcasting commitment to American science, cyber defense, and open-source vulnerability scanning, positioning itself as the trustworthy alternative rather than the efficiency leader. The collective signal is that the market has bifurcated: one track rewards builders who ship products that reduce customer costs, another rewards infrastructure vendors who can prove performance per dollar, and a third rewards labs that can credibly claim alignment with public goods and national priorities. The labs are not competing on the same axis anymore.
Sloane Duvall
A curated reference of models from major AI labs, with open/closed weight status, input modalities, and context window size. American labs tend towards closed weights models and Chinese labs tend toward open weights models.
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None