EN
ZHComplete dining table, price records as well as the resigned V3/R1 endpoints take the new DeepSeek cost webpage. Totally free tier, runs the current V4 models, produced by DeepSeek. We support the model checklist, API prices and you will login tips seemed up against DeepSeek’s individual documents and you can print the brand new remark go out for each web page. To have advice by using the Anthropic API style, please make reference to Anthropic API.
- Register in the chat.deepseek.com — the official cam provides a free of charge level and you may operates the modern V4 patterns.
- Secrets, endpoints, model IDs, fast caching and you can code instances — and just what altered if dated speak and you can reasoner endpoints have been resigned.
- The brand new blended type in rate brings together per provider’s cached and you will uncached type in rate at the cache struck price; team instead of a published cached-type in price is recharged at the full input speed.
- The sooner V3.step 1 and you will R1 models are nevertheless widely used and you can open-lbs, and the dated deepseek-speak and you will deepseek-reasoner API endpoints had been retired to the twenty four July 2026.
- Latest DeepSeek models and you will API costs for each million tokensModelBest forInput (cache miss)OutputGuideDeepSeek-V4.1-FlashFast everyday design.
That is a separate publication, not the state solution. Just before 40 free spins no deposit required powering these types of instances, set the newest DEEPSEEK_API_Secret environment changeable on your terminal to your API secret. If you utilize devices including Claude Password, GitHub Copilot, or OpenCode, you can use DeepSeek as the backend model myself — no code required. The brand new DeepSeek API are backed by of several well-known AI representative and you will coding secretary products.
Comprehend the DeepSeek Utilize Guide for information – 40 free spins no deposit required
The fresh DeepSeek API spends a keen API structure suitable for OpenAI/Anthropic. The price comparator on this page reveals exactly what the same month-to-month workload will cost you for each provider’s individual wrote prices, that have a relationship to the price card. For each million tokens, V4-Flash try $0.15 type in (cache miss) and you will $0.six productivity; V4-Pro try $0.66 enter in and you may $step one.98 output. All of our trial is not the certified service and will not focus on V4. You can also are the brand new totally free demonstration speak on this website, and this operates DeepSeek V3.step one to your area pathways (falling back into R1 whenever a course is actually busy).
Real time changelog out of model launches, prices shifts and you will infrastructure alter. DeepSeek is a Chinese AI lab whose large words habits are strangely inexpensive to work at and open-weight. This can be a non-load example, you might put the new stream parameter to help you true to find weight reaction.
Important factors, endpoints, design IDs, quick caching and you may password advice — as well as just what altered in the event the old cam and reasoner endpoints have been resigned.
The most recent fundamental model is actually DeepSeek-V4.1-Thumb (API term deepseek-flash), essentially available because the 9 September 2026, having native image expertise manufactured in. DeepSeek AI is a good Chinese AI look organization you to makes highest words designs. The state web site is actually deepseek.com as well as the formal speak is actually speak.deepseek.com. All row website links for the provider’s very own price page. Text message and an experimental sight variation.$0.15$0.6Read DeepSeek-V4-ProFlagship reason and you can programming model, to own more difficult multi-action functions.$0.66$1.98Read DeepSeek’s individual speculative decoding pile vessels resources-aware schedulers and you can dos–3× token throughput on the H800s.
Check in from the speak.deepseek.com — the official talk has a totally free level and you may works the modern V4 designs. The fresh blended enter in rates combines for every provider’s cached and you may uncached enter in rate at the cache strike price; company as opposed to a printed cached-type in rates is energized at the complete input rates. Current DeepSeek designs and you can API costs for every million tokensModelBest forInput (cache skip)OutputGuideDeepSeek-V4.1-FlashFast relaxed design. The newest organized provider procedure prompts for the DeepSeek’s system inside Asia, multiple governing bodies limitation they to the certified gadgets, and you can notice-hosting the brand new open loads prevents giving investigation anywhere. The sooner V3.step 1 and R1 habits are still popular and you will discover-lbs, as well as the old deepseek-speak and you will deepseek-reasoner API endpoints have been retired to the twenty four July 2026.
After you’ve received a keen API secret, you have access to the newest DeepSeek model using the following the analogy programs in the OpenAI API style. Distilled versions complement just one 24GB GPU; full patterns you desire multiple-GPU methods. To the wrote API prices DeepSeek is far lesser for every million tokens compared to the GPT, Claude and Gemini flagships, and is aggressive for the programming and you will reasoning benchmarks. Our very own rates web page have an entire dining table, cache-strike rates and you will rate records. Discover weights, tools you desire, and you may step-by-step configurations with Ollama, vLLM and you can llama.cpp. Discover a good workload, then see what a similar frequency costs on every provider’s wrote API costs.