Keys, endpoints, model IDs, quick caching and you may code advice — along with exactly what altered if the old speak and you can reasoner endpoints was resigned. Check in during the chat.deepseek.com — the official chat has a free level and you can runs the present day V4 models. The newest blended type in price combines for every provider’s cached and you will uncached enter in speed at the cache strike speed; business as opposed to a printed cached-type in rates is actually recharged in the full type in price. Newest DeepSeek patterns and API prices for every million tokensModelBest forInput (cache skip)OutputGuideDeepSeek-V4.1-FlashFast casual model. The brand new hosted services process encourages for the DeepSeek’s structure inside China, several governing bodies restrict they for the official gizmos, and you may thinking-holding the brand new discover loads prevents delivering investigation anywhere. The earlier V3.step one and R1 patterns are still commonly used and you may open-pounds, plus the old deepseek-speak and deepseek-reasoner API endpoints have been resigned on the twenty four July 2026.
DeepSeek Use is inside designer preview to own agent utilize builders global. From the altering the brand new configuration, you can utilize the fresh OpenAI/Anthropic SDK otherwise programs appropriate for the fresh OpenAI/Anthropic API to access the newest DeepSeek API. Lookup / Get API is now available, running on all of our in the-house internet search engine and you may support web lookup and you can bring.
Speak assistantRAG / file Q&ACoding agentBulk classification History deepseek-speak casino Yukon Gold sign up and you will deepseek-reasoner aliases was retired; migrate in order to deepseek-v4-flash. V4-Thumb and you may V4-Flash-Vision-Exp is retired; V4-Specialist needs route to V4.1-Thumb out of 14 Sep.
Casino Yukon Gold sign up: DeepSeek patterns and you will what they cost
After you have received an enthusiastic API key, you can access the newest DeepSeek design utilizing the after the analogy programs in the OpenAI API format. Distilled versions match just one 24GB GPU; complete designs you want multi-GPU resources. For the published API costs DeepSeek try much lower per million tokens than the GPT, Claude and you may Gemini flagships, and is competitive to the coding and you can cause benchmarks. Our very own prices page have a complete desk, cache-struck rates and you may rates background. Unlock weights, tools you desire, and you can action-by-step setup that have Ollama, vLLM and you can llama.cpp. Come across a great work, up coming see just what a comparable frequency will cost you on each provider’s wrote API cost.
History API aliases resigned — surcharge still not active
Complete table, rate records as well as the retired V3/R1 endpoints are on the brand new DeepSeek rates webpage. Totally free tier, runs the modern V4 habits, made by DeepSeek. I secure the model number, API prices and log in procedures searched against DeepSeek’s very own documents and print the newest review time on each webpage. To own advice with the Anthropic API structure, please refer to Anthropic API.
Make use of the certified DeepSeek chat
- Discover a workload, next see what an identical regularity costs on every provider’s published API rates.
- Ahead of running these examples, set the newest DEEPSEEK_API_Key environment varying on your terminal for the API trick.
- Open weights, equipment you desire, and you will action-by-step configurations that have Ollama, vLLM and you will llama.cpp.
- For the published API costs DeepSeek is much less for each and every million tokens versus GPT, Claude and you can Gemini flagships, and is competitive to your programming and you will reason criteria.
- Our very own rates page has an entire dining table, cache-strike prices and you will rate records.
- This can be another book, not the official solution.
DeepSeek posts unlock weights for the GitHub and you will Hugging Face, and they work on in your area which have Ollama, vLLM, llama.cpp or LM Facility. This can be a separate book and you may area money. Group discounts, long-perspective surcharges over a great provider’s tolerance, free levels and you may business product sales commonly integrated.
Invoke The new Chat API
The most recent chief design is actually DeepSeek-V4.1-Thumb (API label deepseek-flash), basically available since the 9 September 2026, which have local visualize knowledge built in. DeepSeek AI is an excellent Chinese AI lookup organization one generates high words designs. The official website are deepseek.com as well as the authoritative speak is actually chat.deepseek.com. The line website links to your provider’s individual rates web page. Text and a fresh vision variant.0.150.6Read DeepSeek-V4-ProFlagship need and coding design, to possess more difficult multiple-action functions.0.66step 1.98Read DeepSeek’s individual speculative decoding pile boats tools-aware schedulers and you may dos–3× token throughput to your H800s.
Is actually our totally free trial cam
Live changelog of model launches, rates changes and you will structure transform. DeepSeek try an excellent Chinese AI lab whoever high code models is actually surprisingly cheap to focus on and you can open-weight. That is a non-stream analogy, you could set the new stream parameter so you can real to get stream response.
DeepSeek publishes open loads to your GitHub and you will Hugging Deal with, and so they focus on in your neighborhood having Ollama, vLLM, llama.cpp or LM Business. This is a separate publication and community investment. Group discounts, long-perspective surcharges above a good provider’s endurance, totally free sections and you will firm sale are not incorporated.
DeepSeek Use has become within the developer preview to possess broker use builders around the world. From the altering the fresh arrangement, you need to use the newest OpenAI/Anthropic SDK or software programs suitable for the fresh OpenAI/Anthropic API to get into the newest DeepSeek API. Look / Bring API has become offered, running on the inside-household search and you will help web look and you will get.