OpenAI's newest models promise fewer mistakes and lower inference costs, continuing the same lineage as Astra.
OpenAI released two new models, Sol and Luna, positioning them as cheaper and more accurate successors built on the same foundation as Astra. The launch lands the same week as Anthropic's Opus 5.5, intensifying a pricing and performance race among frontier labs.
Neither model announcement includes new safety commitments beyond standard release notes, though OpenAI has separately said it's bringing in outside mathematicians after recent reputational stumbles tied to overstated math results.
Lower cost, higher accuracy models shift the unit economics for every company building on OpenAI's API — procurement teams should re-run cost models now rather than wait for the next price cut.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →