Cheapest DeepSeek API: up to 99% below list price (November 2026)
Updated November 20, 2026, 11:00 UTC. MartLLM prices are read live when the page is built (every hour); other providers' prices were checked October 30, 2026 or read from their public price lists at build. MartLLM wrote this page and is one of the providers in it.
On November 20, 2026, DeepSeek models on MartLLM cost up to 99% less than their list price, with the card fee added (the best of the 9 DeepSeek models we could price both ways). MartLLM was lowest on both input and output, with the card fee added, for 9 of the 9 DeepSeek models in the table. DeepSeek’s own API charges half price outside its peak hours. MartLLM prices move during the day.
Some MartLLM prices here are more than 70% below list. They are what buyers were billed, not errors. Why so cheap?
MartLLM first because it is lowest on the most models in the table; the rest follow by how they price DeepSeek. An empty cell means we could not confirm it on the provider's own pages.
DeepSeek prices, model by model ($ per 1M input / output tokens)
MartLLM: the best seller’s price, read November 20, 2026; the 5.5% card fee is not in the cell. DeepInfra, Novita and OpenRouter: each one’s public model list, read when this page was built. DeepSeek (peak price, input on a cache miss), Together, Fireworks and Atlas Cloud: their own pricing pages, checked 2026-10-30. An empty cell: not listed.
Pick it if: You want the lowest per-token DeepSeek price now, with the same key for GPT, Claude, Gemini and media models.
Watch for: Market prices move during the day and can rise when the lower-priced sellers are busy (cap them with provider.max_price). Requests go to third-party providers, not to DeepSeek itself.
DeepSeek’s own API charges half price off-peak. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday, excluding Chinese public holidays.
Why are DeepSeek prices on MartLLM so low?
MartLLM is a market: sellers offer DeepSeek capacity, often below DeepSeek’s price, and each request goes to one that can serve it. A request costs what its provider charged, and every request is listed with its cost in Activity.
Can I cap what a DeepSeek request costs on MartLLM?
Yes. Send provider.max_price with the most you will pay per million tokens; the request only goes to sellers at or under it.
Are there other fees on MartLLM?
No MartLLM fee on requests. Adding credit by card costs 5.5% + $0.50.