DeepSeek-V3.2: corrected release date to 1 December 2025
DeepSeek solidifies its 2025 development roadmap with confirmed release date
Corrected release date to 1 December 2025 — verified by the OnAirToday AI Progress Tracker against DeepSeek's published documentation.
OnAirToday
Newsroom
The OnAirToday AI Progress Tracker recorded a verified change to DeepSeek's DeepSeek-V3.2 (version V3.2): Corrected release date to 1 December 2025.
DeepSeek-V3.2 is an open-weight, text-only mixture-of-experts model supporting reasoning, tool use and a documented 128K context window. This profile reflects December 2025 release documentation; current availability, pricing and successor releases could not be checked live.
What changed in this review:
- Replaced zero prices with release-documented USD per million tokens: $0.28 uncached input and $0.42 output; current rates unverified.
- Added published GPQA Diamond, LiveCodeBench v6, Humanity's Last Exam and Terminal-Bench 2.0 results.
- Specified AIME 2025 rather than an unspecified AIME edition.
- Removed the unsupported predecessor-relative price-reduction claim; sparse attention was already introduced in V3.2-Exp.
- Capability score 83.1 → 69
- Flaws and limitations updated
Teams building on DeepSeek-V3.2 should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.
The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/deepseek-v3-2.
References:
- https://huggingface.co/deepseek-ai/DeepSeek-V3.2
- https://arxiv.org/abs/2512.02556
- https://api-docs.deepseek.com/news/news251201
Generated insight by OnAirToday AI
The confirmed December 2025 launch of DeepSeek-V3.2 signals a rapid acceleration in the competitive landscape for open-weights foundational models. This release suggests that DeepSeek is maintaining an aggressive development cadence to challenge established industry incumbents.

Comments
Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.