ChatGPT o3-mini is (probably) out tomorrow, so we should talk about benchmarks

Source: Nate’s Newsletter Date: 2025-01-30 URL: https://natesnewsletter.substack.com/p/chatgpt-o3-mini-is-probably-out-tomorrow

Summary

Paywalled preview only. Nate anticipates o3-mini landing the following day, framing it as meaningfully better than o1 while being significantly faster and cheaper — a performance-per-dollar step change. The promised argument about benchmark methodology (what the headline flags) is gated.

Implications

AI economics thread. If o3-mini delivered the cost/performance ratio Nate projected, it resets the baseline for what “affordable reasoning” means — compressing the price floor for inference-heavy workloads.

Vendor positioning thread. A cheaper, faster o-series model at launch would intensify pressure on Anthropic’s Sonnet-tier and Google’s Flash-tier pricing. OpenAI using mini models to dominate the cost-sensitive segment while protecting frontier pricing is a deliberate two-tier strategy.

Watch: Actual o3-mini pricing and throughput at launch vs. o1 — the delta there is what moves enterprise adoption curves.