June: flashy ML-perf demo with CoreWeave using 22k H100s
November: actually trained model using 5k H100s
June: we claim Inflection-1 is the best model in its compute class and are preparing a frontier model
November: we beat PaLM 2, which everyone else forgot about long ago anyway
Inflection got a ton of hype with its $1.3B raise (likely not cash but principally GPU compute credits) earlier this year, but now is starting to look like the next victim of inflated expectations.
No comments yet.
