Sobre DeepSeek
DeepSeek is an AI company operated by Hangzhou DeepSeek Artificial Intelligence Co., Ltd. Their mission is to develop open-weight large language models that enable reasoning, tool-use and private deployment, offering an alternative path in the large-model ecosystem.
Desempenho
Desempenho recente do Gateway nos modelos de DeepSeek.
Modelos recentes
Lançamentos recentes com detalhes de disponibilidade, contexto, preços e uso.
DeepSeek V4.1 Flash 1017 is the 17 October 2026 update with improved reasoning, coding, agentic and visual performance, shorter task execution time and output length, native image understanding, a 1M-token context window and up to 384K output tokens.
DeepSeek V4.1 Flash is a 552B-backbone multimodal Mixture-of-Experts model with 8B active parameters during prefill and 16B during decode, a 1M-token context window, native image understanding, continuously controllable reasoning effort, and optimized agentic performance.
DeepSeek V4 Flash Vision Exp is DeepSeek's experimental multimodal vision-understanding model. It combines V4 Flash-level text, reasoning, and agent capabilities with image understanding.
DeepSeek V4 Pro 0813 is the 13 August 2026 version served by DeepSeek's stable deepseek-v4-pro API slug. It supports thinking and non-thinking modes, JSON output, tool calls, Chat Completions, Responses, Anthropic Messages, prefix completion, fill-in-the-middle completion, a 1M-token context window, and up to 384K output tokens. DeepSeek postponed discontinuation of the V4 Pro service to 12:00 Beijing Time on 14 September 2026; a future V4.1 Pro is intended to replace it.
DeepSeek V4 Flash 0731 is the official 31 July 2026 release that supersedes the V4 Flash preview, with substantially stronger agent capabilities and a DSpark speculative-decoding module. It supports native Responses API use, tool use, structured output, server-side web search, a 1M-token context window, and up to 384K output tokens.
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window.
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window.
DeepSeek V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance.
DeepSeek V3.2 Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It accepts text inputs and produces text outputs.
DeepSeek V3.2 Thinking is a reasoning model from DeepSeek for multi-step analysis, math, and tool-driven tasks. It accepts text inputs and produces text outputs.
DeepSeek OCR is a document understanding model from DeepSeek for OCR, extraction, and page-level visual reasoning. It is suited to extracting text, structure, and answers from PDFs, scans, and screenshots.
DeepSeek V3.2 Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. This entry tracks a preview release and capabilities may change before general availability.
Na web
Perfis oficiais, pesquisas e canais da comunidade.
