Very first API Label – 50 free spins on tiki fruits DeepSeek API Docs

Important factors, endpoints, design IDs, quick caching and you may password instances — along with exactly what altered when the old chat and reasoner endpoints were retired. Sign in from the speak.deepseek.com — the state talk has a free level and works the current V4 patterns. The newest combined input price mixes for every provider’s cached and uncached input rate at your cache hit price; organization instead of a printed cached-input rate try recharged in the full type in price. Current DeepSeek designs and you can API cost for each and every million tokensModelBest forInput (cache miss)OutputGuideDeepSeek-V4.1-FlashFast casual design. The fresh managed services processes encourages to the DeepSeek’s infrastructure in the Asia, multiple governments limit they to the formal gadgets, and you will notice-hosting the brand new unlock weights stops giving research anywhere. The earlier V3.step 1 and you may R1 models are still commonly used and discover-weight, and also the dated deepseek-speak and you can deepseek-reasoner API endpoints was retired on the twenty-four July 2026.

DeepSeek Funnel is becoming within the designer examine to possess broker funnel developers worldwide. From the altering the brand new arrangement, you need to use the fresh OpenAI/Anthropic SDK or softwares appropriate for the brand new OpenAI/Anthropic API to access the fresh DeepSeek API. Search / Bring API has become readily available, running on our very own in the-family search and you will supporting web research and get.

Speak assistantRAG / file Q&ACoding agentBulk classification History deepseek-speak and deepseek-reasoner aliases have been retired; migrate so you can deepseek-v4-thumb. V4-Flash and you will V4-Flash-Vision-Exp try retired; V4-Pro demands path to V4.1-Flash away from 14 September.

DeepSeek models and whatever they rates – 50 free spins on tiki fruits

Once you have received an enthusiastic API secret, you have access to the newest DeepSeek model by using the following the analogy scripts on the OpenAI API structure. Distilled versions complement a single 24GB GPU; full models you need multiple-GPU equipment. On the composed API costs DeepSeek are far lesser for every million tokens than the GPT, Claude and you may Gemini flagships, and is also competitive for the coding and you can reason standards. Our very own cost web page provides an entire table, cache-strike prices and you will speed records. Open weights, methods you desire, and you will step-by-step configurations having Ollama, vLLM and you can llama.cpp. Discover a great work, then see just what an identical regularity will cost you on every provider’s authored API costs.

History API aliases resigned — surcharge however maybe not energetic

Complete desk, rates background and the 50 free spins on tiki fruits resigned V3/R1 endpoints are on the new DeepSeek rates webpage. 100 percent free level, works the modern V4 models, created by DeepSeek. I support the model listing, API cost and you may log in tips appeared up against DeepSeek’s own records and you will printing the fresh remark time on every web page. To own examples with the Anthropic API style, excite consider Anthropic API.

Make use of the formal DeepSeek cam

Come across an excellent workload, following see just what a comparable volume will cost you on each provider’s published API cost.
Prior to powering these instances, lay the new DEEPSEEK_API_Trick environment changeable on the terminal to your API trick.
Discover loads, equipment you need, and you may action-by-step settings that have Ollama, vLLM and you may llama.cpp.
For the authored API rates DeepSeek is 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 criteria.
All of our prices webpage provides an entire table, cache-struck costs and you will speed background.
That is a different publication, not the official service.

DeepSeek posts unlock weights to the GitHub and Hugging Face, and they focus on in your town that have Ollama, vLLM, llama.cpp or LM Business. This really is an independent guide and you may neighborhood money. Batch discounts, long-perspective surcharges above a good provider’s endurance, totally free levels and business sale are not integrated.

Invoke The new Cam API

Its latest main design are DeepSeek-V4.1-Flash (API identity deepseek-flash), generally readily available because the 9 Sep 2026, which have native photo expertise manufactured in. DeepSeek AI are an excellent Chinese AI look team you to produces highest language models. The state site is deepseek.com as well as the authoritative speak is actually speak.deepseek.com. All of the row website links for the provider’s individual rate page. Text message and a fresh attention variant.0.150.6Read DeepSeek-V4-ProFlagship reason and you will programming design, to possess more difficult multiple-step work.0.661.98Read DeepSeek’s individual speculative decoding pile ships tools-aware schedulers and dos–3× token throughput for the H800s.

Are all of our 100 percent free demonstration talk

Live changelog from design releases, cost shifts and you may structure alter. DeepSeek try an excellent Chinese AI lab whose high language models is strangely inexpensive to focus on and you will open-lbs. This really is a low-weight example, you can put the brand new stream parameter in order to real discover weight effect.

DeepSeek posts open loads for the GitHub and you may Hugging Deal with, and so they focus on in your neighborhood which have Ollama, vLLM, llama.cpp otherwise LM Studio. This can be an independent publication and community funding. Group offers, long-context surcharges a lot more than a good provider’s threshold, free levels and you can corporation sale aren’t integrated.