Frontier Log

Frontier Log positions itself as a structured record of AI progress rather than a news outlet. Instead of commentary, it collects dated events — model launches, policy decisions, funding rounds, safety assessments and stealth releases — and ties each one to the official post from the lab that announced it. The result is a chronological reference you can cite rather than a stream of opinion.

The site is organized around four views: a timeline of events, a model directory, benchmark scores and narrative stories. Tags such as Open Weights, Small Models and Enterprise let you filter across providers including OpenAI, Anthropic, Google, NVIDIA, Moonshot AI, MiniMax and DeepSeek. A free API, usable without a key, exposes the same data for programmatic access.

It is most useful for AI researchers, policy analysts, product managers and developers who need to verify when something shipped or how a model scored. Journalists and investors can use the timeline to trace how announcements cluster, while engineers can pull the API into internal tracking tools.

Because entries are limited to official sources, the log is a strong starting point but not a complete picture — independent evaluations and smaller labs may be underrepresented. Cross-check benchmark figures against the original posts before citing them, and treat the stories section as context rather than primary evidence.