DeepSeek V3 moves to version DeepSeek-V3 (original December 2024 release)

The OnAirToday AI Progress Tracker recorded a verified change to DeepSeek's DeepSeek V3 (version DeepSeek-V3 (original December 2024 release)): Version now DeepSeek-V3 (original December 2024 release) (was deepseek-chat).

DeepSeek-V3 is an open-weight, text-focused mixture-of-experts model with 671 billion total parameters and 37 billion activated per token. Its original release must be distinguished from later V3-family models and the mutable deepseek-chat API alias; current original-model API pricing is Not yet confirmed by public reporting.

What changed in this review:

• Corrected original DeepSeek-V3 release timing to 26 December 2024 and context window to 128K tokens.
• Separated original DeepSeek-V3 from later V3.2 releases and the mutable deepseek-chat API alias.
• Replaced benchmark figures with original DeepSeek-V3 results published by DeepSeek.
• Current original-model API prices are null; this does not establish a price change.
• Added attributed family-level misuse and harness-exploit coverage without treating it as confirmed original-V3 exposure.
• Related distillation reports do not establish separate underlying incidents; incident-count articles do not identify additional cases.

Teams building on DeepSeek V3 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-chat.

References:

• https://github.com/deepseek-ai/DeepSeek-V3
• https://arxiv.org/abs/2412.19437
• https://api-docs.deepseek.com/news/news1226
• https://huggingface.co/deepseek-ai/DeepSeek-V3.2