DeepSeek: DeepSeek V3

DeepSeek V3 is a text-only model from DeepSeek with a 131,072-token context window and a 16,000-token output ceiling. It accepts text input only, supports tool use, and does not include a built-in reasoning mode. Structured output support is unconfirmed from available data. At $0.20 per million input tokens and $0.80 per million output tokens, DeepSeek V3 sits at the low end of the pricing range for capable general-purpose models. Its blended benchmark score of 79.1 across four benchmarks is modest in coverage, so treat it as directional rather than definitive; the standout figure is a 97.9 on AI Index coding tasks, which makes it worth shortlisting for code-heavy workflows on a budget. Teams running agentic pipelines should note the 60.0 agentic score and weigh that against alternatives with stronger tool-orchestration track records.