Models & pricing
DarkSeek token prices are listed in USD per 1M tokens. Input and output tokens have separate rates. The same prices apply at all hours, every day: there is no peak-hour surcharge.
此链接保留对应的中文文档路径。目前技术参考内容为英文；功能可用性以 DarkSeek 的实际部署和验证状态为准。
Model details#
DarkSeek uses the model identifier darkseek-v4.1. There is no DarkSeek Pro tier. The intended maximum output is 384K tokens, matching the reference product. This is a release target: public 384K output is not yet certified, and preview deployments may enforce a lower ceiling. Query the deployed Models response for its actual max_output_tokens.
| MODEL | DarkSeek |
|---|---|
| API model identifier | darkseek-v4.1 |
| Base URL (OpenAI format) | https://api.darkseeks.com/v1 |
| Service status | Developer preview; public inference remains unverified |
| Context length | Model server reports 1,048,576 tokens; public long-context certification pending |
| Max output | MAXIMUM: 384K — release target; public deployment certification pending |
| Default output | 4,096 tokens unless configured otherwise |
Features#
Implementation and verification status are shown separately from public availability. Protocol tests do not certify every model behavior or make an endpoint publicly available.
| FEATURE | DARKSEEK STATUS |
|---|---|
| JSON Output | Preview — chat request forwarding; validate returned JSON |
| Tool Calls | Preview — protocol support and runtime parsers; public certification pending |
| Responses API | Implemented and SDK/protocol tested; public launch pending |
| Anthropic API | Messages implemented and SDK/protocol tested; public launch pending |
| Chat Prefix Completion (Beta) | Not certified |
| FIM Completion (Beta) | Not certified; suffix requests are rejected |
| Vision | Not enabled in the public gateway; text-only input |
Token pricing#
One flat rate, 24/7. The same DarkSeek prices apply at all times, without a peak/off-peak timetable. All prices below are per 1,000,000 tokens.
| TOKEN TYPE | PRICE / 1M TOKENS (USD) |
|---|---|
| Input tokens (cache hit) | $0.003 |
| Input tokens (cache miss) | $0.15 |
| Output tokens | $0.60 |
Usage calculation#
At these rates, token cost in USD = (cache-hit input tokens × 0.003 + cache-miss input tokens × 0.15 + output tokens × 0.60) / 1,000,000. For example, 1M cache-miss input tokens and 1M output tokens total $0.75. Cached-input pricing requires measured cache-hit usage; repeated prompts or faster responses are not proof of a cache hit.
Search & Fetch pricing#
Search and Fetch are not enabled for public use, and no DarkSeek per-request price is published for them. The token prices above do not advertise availability of those APIs.
Billing and deduction rules#
This is the published DarkSeek token price schedule. Automated token billing, prepaid payments and balance deductions are not enabled yet. Publishing prices does not activate checkout or charge an account. Actual charges require measured token usage, reconciled accounting and verified payment settlement. No sample balance or estimated usage is treated as a real bill.