Sarashina3.2 mini thinking launches with API access

Sarashina's significance is the addition of reasoning capability to a Japanese-focused model family, rather than evidence of a new global frontier. SB Intuitions says the model was built from scratch and adds thinking before responding. My assessment is that this matters because domestic model builders need a path beyond fluent local-language chat toward systems that can execute useful work. An accessible API gives developers a practical way to investigate that transition without building the underlying model themselves. ([sbintuitions.co.jp](https://www.sbintuitions.co.jp/blog/entry/2026/10/29/133000?utm_source=openai))

The three reasoning settings make the competitive question concrete: does extra thinking produce enough additional task success to justify its latency? Tool calling and search are particularly important tests because useful agents must interact with systems outside their own generated text. For Race to AGI readers, this is a regional capability signal worth tracking, not a basis for changing an AGI forecast. The next evidence should come from independent Japanese-language workflows measuring completion rates, costs and failure modes, rather than headline benchmark comparisons. A stronger result there would establish a clearer commercial role for this model. ([itmedia.co.jp](https://www.itmedia.co.jp/aiplus/article/2610/30/2000002743/))