What changed
DeepSeek released V3.2 with thinking mode that could interleave reasoning and tool calls, plus a temporary V3.2-Speciale endpoint for a higher-compute reasoning variant.
Why it matters
Reasoning and action were becoming one loop rather than separate planning and execution calls, making tool traces more capable but also harder to govern.
What to do
Use the standard model for tool-rich research and coding; log every tool decision and set cost, recursion, and authority limits before enabling thinking mode in production.
Caveat
DeepSeek's benchmark gains were vendor-reported, and the Speciale API endpoint was temporary through 15 December 2025 rather than a durable production offering.