Important factors, endpoints, model IDs, punctual caching and you may code examples — in addition to what changed if old chat and reasoner endpoints was resigned. Register during the chat.deepseek.com — the official speak features a free tier and works the present day V4 patterns. The brand new combined type in speed brings together for each provider’s cached and uncached input rates at your cache strike speed; business instead a published cached-type in speed is recharged at the complete enter in rate. Current DeepSeek models and you can API rates for each and every million tokensModelBest forInput (cache miss)OutputGuideDeepSeek-V4.1-FlashFast relaxed model. The newest hosted service techniques prompts to your DeepSeek’s system inside Asia, numerous governing bodies limit it to the authoritative products, and you will thinking-holding the newest discover loads stops sending research everywhere. The earlier V3.step one and you can R1 designs are still widely used and you will open-lbs, as well as the dated deepseek-speak and deepseek-reasoner API endpoints have been resigned for the twenty four July 2026.
DeepSeek Harness has become in the designer examine for representative funnel developers worldwide. Because of the altering the fresh arrangement, you can use the newest OpenAI/Anthropic SDK otherwise softwares suitable for the newest OpenAI/Anthropic API to get into the newest DeepSeek API. Research / Bring API has become available, running on all of our in the-home search and you may supporting net search and you can get.
Chat assistantRAG / document Q&ACoding agentBulk category History deepseek-cam and you can deepseek-reasoner aliases have been retired; migrate in order to deepseek-v4-flash. V4-Thumb and you may V4-Flash-Vision-Exp is actually retired; V4-Pro needs path to V4.1-Flash from 14 Sep.
Rainbow riches symbols | DeepSeek models and you will what they rates
After you have gotten an enthusiastic API trick, you have access to the newest DeepSeek design utilizing rainbow riches symbols the pursuing the analogy programs in the OpenAI API format. Distilled alternatives fit just one 24GB GPU; complete habits you would like multi-GPU equipment. For the authored API costs DeepSeek is actually much lower for each and every million tokens compared to the GPT, Claude and you will Gemini flagships, and it is competitive for the programming and you may reasoning standards. The prices page provides a full dining table, cache-struck rates and price records. Unlock weights, methods you need, and you can action-by-step setup that have Ollama, vLLM and you can llama.cpp. Discover a good workload, next see what an identical volume will cost you on each provider’s wrote API costs.
History API aliases resigned — surcharge still maybe not active
Full table, price background and also the resigned V3/R1 endpoints are on the fresh DeepSeek costs web page. Totally free level, operates the present day V4 models, created by DeepSeek. We contain the model list, API prices and you can sign on actions appeared against DeepSeek’s individual paperwork and you may printing the fresh opinion time on every web page. To have instances by using the Anthropic API structure, excite consider Anthropic API.
Use the formal DeepSeek cam
- See a great work, then see just what the same volume will cost you on each provider’s published API rates.
- Prior to powering such instances, put the fresh DEEPSEEK_API_Trick ecosystem adjustable on your own critical to your API secret.
- Unlock weights, equipment you want, and step-by-step setup with Ollama, vLLM and you can llama.cpp.
- On the published API prices DeepSeek is much smaller for each and every million tokens versus GPT, Claude and you may Gemini flagships, and is aggressive to the coding and you will reason benchmarks.
- Our very own prices page have a full table, cache-strike prices and you will rate background.
- This can be another guide, perhaps not the official provider.
DeepSeek publishes open loads for the GitHub and you may Hugging Deal with, plus they work on in your neighborhood that have Ollama, vLLM, llama.cpp or LM Business. This can be another book and area financing. Batch savings, long-perspective surcharges above a great provider’s tolerance, free levels and business selling aren’t integrated.
Invoke The fresh Cam API
Their current head model are DeepSeek-V4.1-Thumb (API name deepseek-flash), generally offered because the 9 September 2026, that have indigenous visualize expertise produced in. DeepSeek AI is an excellent Chinese AI lookup company you to makes high language models. The official webpages is deepseek.com plus the formal speak is actually talk.deepseek.com. The row hyperlinks for the provider’s own price webpage. Text message as well as an experimental vision variant.0.150.6Read DeepSeek-V4-ProFlagship reason and you will coding model, to have more challenging multi-step performs.0.661.98Read DeepSeek’s own speculative decryption heap vessels resources-aware schedulers and you can 2–3× token throughput to your H800s.
Is all of our totally free demo chat
Alive changelog out of design launches, cost shifts and you can system transform. DeepSeek are a great Chinese AI research whoever high words patterns try strangely cost effective to focus on and open-lbs. That is a low-load example, you could lay the newest weight factor so you can true to locate load effect.
DeepSeek posts discover loads to your GitHub and Hugging Face, and so they focus on locally that have Ollama, vLLM, llama.cpp or LM Studio. This can be another guide and area financing. Group discounts, long-framework surcharges more than a great provider’s tolerance, 100 percent free sections and business selling commonly incorporated.
DeepSeek Funnel is inside the designer preview to have broker harness developers worldwide. By changing the new arrangement, you can utilize the brand new OpenAI/Anthropic SDK or programs appropriate for the newest OpenAI/Anthropic API to view the fresh DeepSeek API. Look / Fetch API is offered, running on our very own in the-family search engine and help web look and you may fetch.
