DeepSeek is a Chinese AI company that builds open-weight foundation models and offers a hosted API compatible with OpenAI and Anthropic clients. Its V4 models target long-context reasoning, coding and agentic workloads at low token cost.

Which DeepSeek models can I call through Felo?

Felo currently lists DeepSeek V4.1 Flash, DeepSeek V4.1 Flash 1017 and DeepSeek V4 Pro 0813. The live table on this page is the authoritative list of routes and rates.

Do I need a DeepSeek account or API key?

No. DeepSeek models on Felo use a Felo API key and are billed from your Felo balance, so no DeepSeek account, top-up or key is required.

How is DeepSeek billed on Felo?

Per token in USD at the rate published for each route, with cached input priced separately where supported. DeepSeek's direct API charges less off-peak and prices cache hits separately, but Felo requests follow the rate shown in Felo's model table.

Are DeepSeek's models open weights?

Yes. DeepSeek publishes V4 model weights on Hugging Face under the MIT license, so the same models can be self-hosted or used through the hosted API. Felo provides a managed route without needing to run the weights yourself.

What is DeepSeek best for?

Input-heavy, long-context work such as codebase and log analysis, document processing and agentic tool use, especially where token cost matters. DeepSeek routes on Felo return text, so image, video and audio generation need a different provider.

DeepSeek API

DeepSeek is a Chinese AI company (Hangzhou DeepSeek AI) known for open-weight models and low-cost inference; its V4 family is built for long-context reasoning, coding and agentic work. DeepSeek exposes the models through its own API, which accepts OpenAI- and Anthropic-compatible requests, and publishes V4 weights for self-hosting. Through Felo, three DeepSeek routes are available under one Felo key and balance, with no DeepSeek account required.

Official website

Models: 3; chat ×3
Input per 1M: $0.3–$1.32
Output per 1M: $1.2–$3.96
Largest context: 1.049M

What DeepSeek provides

DeepSeek builds foundation models and publishes them as open weights alongside a hosted API. Its V4 generation uses a sparse mixture-of-experts architecture with a 1M-token context window and a 384K output limit; the API supports thinking and non-thinking modes, JSON output, tool calls, the OpenAI Responses format and an Anthropic-compatible endpoint. DeepSeek also publishes model weights under the MIT license on Hugging Face, so teams can self-host instead of using the hosted API.

DeepSeek models available on Felo

Felo currently lists three DeepSeek routes: DeepSeek V4.1 Flash, DeepSeek V4.1 Flash 1017 and DeepSeek V4 Pro 0813. V4.1 Flash is the multimodal member, accepting text and images and returning text, while V4 Pro 0813 is the larger text-only option; all three carry a 1M-token context window. DeepSeek's own API has already begun routing deepseek-v4-pro requests to V4.1 Flash, so the Felo table is the authoritative place to see which DeepSeek routes and prices are live.

Pricing and access

On Felo, DeepSeek usage is billed per token in USD from your Felo balance. Published default-channel rates for the three routes currently range from $0.30 to $1.32 per 1M input tokens and from $1.20 to $3.96 per 1M output tokens, with cached input billed at a lower rate. DeepSeek's direct API prices by peak and off-peak windows, charging half the peak rate off-peak, and bills cache hits at a separate lower price; on Felo, the model table publishes the rate that applies to each route, including cache pricing where the route supports it. No DeepSeek account is required.

Why teams use DeepSeek through Felo

DeepSeek's models are aimed at input-heavy workloads where cost is dominated by large codebases, logs and documents, and at teams that want an open-weight fallback if a hosted route changes. On Felo, DeepSeek sits next to Claude, Gemini and other providers on one key, which makes it practical to route high-volume or long-context traffic to DeepSeek while keeping a frontier model for harder work. Billing and model comparison stay in one account.

Limits to know

Felo lists DeepSeek text routes only; DeepSeek does not offer image, video or audio generation in this catalog, and only the Flash models accept image input. DeepSeek's own API applies concurrency limits and time-of-day pricing, and its model lineup changes quickly: V4 Pro is being phased out in favor of V4.1 Flash, and the 1017 revision supersedes the original V4.1 Flash. Check the live table for the current route set, and the model page for context and output limits.

DeepSeek V4.1 Flash 1017

Input per 1M: $0.3
Output per 1M: $1.2
Context: 1.049M
Max output: 393.216K

DeepSeek V4.1 Flash

Input per 1M: $0.3
Output per 1M: $1.2
Context: 1.049M
Max output: 131.072K

DeepSeek V4 Pro 0813

Input per 1M: $1.32
Output per 1M: $3.96
Context: 1.049M
Max output: 393.216K

DeepSeek models compared by input price, output price, context window and maximum output.
Model	Input per 1M	Output per 1M	Context	Max output
DeepSeek V4.1 Flash 1017	$0.3	$1.2	1.049M	393.216K
DeepSeek V4.1 Flash	$0.3	$1.2	1.049M	131.072K
DeepSeek V4 Pro 0813	$1.32	$3.96	1.049M	393.216K

Prices are published standard rates in USD per 1M tokens, before channel discounts.

Frequently asked questions

What is DeepSeek?: DeepSeek is a Chinese AI company that builds open-weight foundation models and offers a hosted API compatible with OpenAI and Anthropic clients. Its V4 models target long-context reasoning, coding and agentic workloads at low token cost.
Which DeepSeek models can I call through Felo?: Felo currently lists DeepSeek V4.1 Flash, DeepSeek V4.1 Flash 1017 and DeepSeek V4 Pro 0813. The live table on this page is the authoritative list of routes and rates.
Do I need a DeepSeek account or API key?: No. DeepSeek models on Felo use a Felo API key and are billed from your Felo balance, so no DeepSeek account, top-up or key is required.
How is DeepSeek billed on Felo?: Per token in USD at the rate published for each route, with cached input priced separately where supported. DeepSeek's direct API charges less off-peak and prices cache hits separately, but Felo requests follow the rate shown in Felo's model table.
Are DeepSeek's models open weights?: Yes. DeepSeek publishes V4 model weights on Hugging Face under the MIT license, so the same models can be self-hosted or used through the hosted API. Felo provides a managed route without needing to run the weights yourself.
What is DeepSeek best for?: Input-heavy, long-context work such as codebase and log analysis, document processing and agentic tool use, especially where token cost matters. DeepSeek routes on Felo return text, so image, video and audio generation need a different provider.