HTTP 200 verified twice
Finding your next stop…
Loading the latest directory information.
Loading the latest directory information.
DeepSeek (Hangzhou DeepSeek AI) is a Chinese AI lab building open-source large models under MIT License, including V4, V3, and R1 series. It offers web, app, and API access with OpenAI/Anthropic compatibility, 1M-token context, reasoning modes, and tool use.
Start here for the decision-making essentials: what DeepSeek does, who it is for, how it is accessed, and the first-party sources behind this profile.
0.5 yuan (cache hit) / 2 yuan (cache miss) per million input tokens
8 yuan per million ou
Read-only public captures of DeepSeek’s homepage. Screenshots are dated, never live embeds, and open full-screen.
Short answers to the questions buyers and builders commonly ask about DeepSeek. Each answer cites the shared ledger below, where every source is listed once.
deepseek-chat 与 deepseek-reasoner 将于 2026-07-24 停止使用
旧有的 API 接口的两个模型名 deepseek-chat 与 deepseek-reasoner 将于三个月后（2026-07-24）停止使用
0.5 yuan (cache hit) / 2 yuan (cache miss) per million input tokens; 8 yuan per million ou
每百万输入 tokens 0.5 元（缓存命中）/ 2 元（缓存未命中），每百万输出 tokens 8 元
支持 OpenAI ChatCompletions 接口与 Anthropic 接口 · Supports Anthropic API format; endpoints deepseek-chat (non-thinking) and deepseek-reasone
支持 OpenAI ChatCompletions 接口与 Anthropic 接口
SGLang and LMDeploy for FP8 inference; TensorRT-LLM and MindIE for BF16 inference
SGLang 和 LMDeploy 第一时间支持了 V3 模型的原生 FP8 推理，同时 TensorRT-LLM 和 MindIE 则实现了 BF16 推理
开源 DeepSeek-V4、DeepSeek-R1 等前沿大模型 · Base and post-training weights released on Hugging Face and ModelScope
开源 DeepSeek-V4、DeepSeek-R1 等前沿大模型
This profile connects the jobs DeepSeek is described as handling with its delivery model, access options and the subjects used to match it to related products in this directory.
Each fact points to a recorded source, making it easy to distinguish verified product information from claims that need checking.
HTTP 200 verified twice
DeepSeek | 深度求索
Company: 杭州深度求索人工智能基础技术研究有限公司 · Mission focuses on world-leading general AI foundation models and techniques · Open-sources DeepSeek-V4, DeepSeek-R1 and other frontier large models
杭州深度求索人工智能基础技术研究有限公司
专注于研究世界领先的通用人工智能底层模型与技术
前沿大模型（如 DeepSeek-V4、DeepSeek-R1 系列）以开源仓库及模型权重的形式公开
MIT License
DeepSeek V4, V3.2, V3.1, R1, V3
DeepSeek-V4 最大上下文长度为 1M（百万字）
支持 OpenAI ChatCompletions 接口与 Anthropic 接口
DeepSeek-R1-0528 支持工具调用
支持 Responses API 和 Codex 接入
提供网页版、官方 App 和 API 开放平台
A concise view of the jobs, capabilities and integrations described in the recorded product sources.
Documented product formats, platforms and official distribution destinations. Availability can vary by region and plan.
A concise history of software releases and material product changes. Events appear only when a dated source supports what changed.
Showing the newest updates and meaningful milestones. Open an entry for its summary and source.
随 V4 上线，官方公告旧 API 模型名 deepseek-chat 与 deepseek-reasoner 将于三个月后（2026-07-24）停止使用，过渡期分别指向 V4-Flash 的非思考模式与思考模式。
View source [3]DeepSeek released the preview of its DeepSeek-V4 model series with 1M-token context, available as V4-Pro and V4-Flash via web, app, and API. The release introduces DSA sparse attention and supports both OpenAI and Anthropic API formats.
View source [3]DeepSeek released V3.2 and V3.2-Speciale as stable versions, with V3.2 being the first model to integrate thinking into tool use. V3.2 became the default model on web, app, and API, with V3.2-Speciale available as a temporary research API.
View source [6]DeepSeek released V3.1 with a hybrid reasoning architecture supporting both thinking and non-thinking modes in one model, expanded context to 128K, and added Anthropic API format support. API pricing changes were scheduled for September 6, 2025.
View source [5]DeepSeek released R1-0528, a minor update with deeper reasoning, hallucination rate reduced by 45-50%, and tool calling support. The model weights and a distilled Qwen3-8B variant were open-sourced under MIT License.
View source [4]Facts, answers, structured details, milestones and primary resource links cite this shared ledger. Each external page appears once; release tags from the same GitHub project are grouped under one release history.
These tools share a workflow, capability or audience with DeepSeek. They may complement it rather than replace it, and are not presented as integrations or endorsements.
deepseek-chat 与 deepseek-reasoner 将于 2026-07-24 停止使用
支持 JsonOutput
DeepSeek-V3
671B MoE parameters with 37B activated per token
Mixture-of-Experts (MoE), self-developed by DeepSeek
14.8T tokens of pre-training data
Native FP8 weights with BF16 conversion scripts provided
Base and post-training weights released on Hugging Face and ModelScope
SGLang and LMDeploy for FP8 inference; TensorRT-LLM and MindIE for BF16 inference
0.5 yuan (cache hit) / 2 yuan (cache miss) per million input tokens; 8 yuan per million output tokens
Text-only — DeepSeek-V3 does not support multimodal input or output
Supports Anthropic API format; endpoints deepseek-chat (non-thinking) and deepseek-reasoner (thinking)
DeepSeek-V3.2 is the first DeepSeek model to integrate thinking into tool use
Company: 杭州深度求索人工智能基础技术研究有限公司 (Hangzhou DeepSeek AI) · Mission: focused on world-leading AGI foundational models and technology · Open-source frontier models include DeepSeek-V4 and DeepSeek-R1 series
杭州深度求索人工智能基础技术研究有限公司 (Hangzhou DeepSeek AI)
专注于研究世界领先的通用人工智能底层模型与技术，目标是实现 AGI
MIT License（开源仓库及模型权重）
V4-Pro 与 V4-Flash 支持 OpenAI ChatCompletions 接口与 Anthropic 接口，base_url 不变，model 参数为 deepseek-v4-pro 或 deepseek-v4-flash
V4 系列官方服务的标准上下文长度为 1M（一百万）tokens
DeepSeek-R1-0528 开源版本上下文长度为 128K（网页端、App 和 API 提供 64K 上下文）
V4-Pro 与 V4-Flash 同时支持非思考模式与思考模式；思考模式支持 reasoning_effort 参数设置思考强度（high/max）
DeepSeek-R1-0528 支持工具调用（Tau-Bench：airline 53.5% / retail 63.9%），但不支持在 thinking 中进行工具调用
R1-0528 API 新增 Function Calling 和 JsonOutput 支持
DeepSeek-R1-0528 模型参数为 685B（其中 14B 为 MTP 层）
DeepSeek-V3 is a 671B MoE model with 37B activated parameters, pre-trained on 14.8T tokens
DeepSeek-V3 weights are open-sourced on Hugging Face in native FP8, with FP8-to-BF16 conversion scripts
DeepSeek-V3.2 and DeepSeek-V3.2-Speciale are open-sourced on HuggingFace and ModelScope
