OpenAI is doubling down on infrastructure and operational security while the rest of the field splinters across use cases and hardware optimization. OpenAI's two announcements, one on serving ChatGPT at billion-user scale through Habitat's multi-tenancy architecture, the other on containment for frontier model training, signal a company focused on the unglamorous work of running production systems and hardening internal processes at a scale competitors haven't reached. Meanwhile, Google is positioning generative AI as a tool for specific domains like tuberculosis screening, Hugging Face is building developer-facing utilities for model discovery and hardware compatibility, AMD is publishing optimization guides for 3D graphics and matrix operations on its chips, Anthropic is conducting security assessments, and Moonshot is releasing Kimi K3.1 Pro. The pattern is clear: OpenAI treats infrastructure and security as competitive moats; everyone else is either solving narrower problems, enabling others' workloads, or publishing technical guides to commoditize hardware advantages. The infrastructure plays matter more than the model releases because they determine who can actually serve users at volume without melting their cost structure.
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