Using the DeepSeek API outside China: four ways, with prices
You can use DeepSeek's models on DeepSeek's own platform, through a host such as DeepInfra, through a router such as OpenRouter, or through a multi-model gateway such as XHuoAPI. Below are the public prices for the same models in each place, each with its source and the date we checked it. XHuoAPI is this site; we list our price in all four tiers.
Third-party prices checked:
DeepSeek V4.1 Flash
DeepSeek now uses the name deepseek-flash for V4.1 Flash; the old name deepseek-v4-flash is still accepted and billed at the Flash price.
| Where | Model ID | $ / 1M in / out | Notes |
|---|---|---|---|
| DeepSeek (official) source | deepseek-flash | $0.30 / $1.20 | Peak price. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday; all other hours (including weekends) are half price: $0.15 / $0.60 |
| DeepInfra source | deepseek-ai/DeepSeek-V4.1-Flash | $0.14 / $0.42 | Promotional price, fp8 precision. Without the promotion: $0.20 / $0.60 |
| OpenRouter source | deepseek/deepseek-v4.1-flash-1017 | $0.30 / $1.20 | The price shown is DeepSeek's own endpoint on OpenRouter. OpenRouter routes across providers; this model has 33, and the lowest in / out price is $0.070 / $0.15, mostly from fp4 / fp8 quantized hosts. |
| XHuoAPI (Standard) | deepseek-v4.1-flash | $0.24 / $0.96 | Saver $0.15 / $0.60 Spot $0.090 / $0.36 Official tier $0.30 / $1.20 The same key also calls Qwen, GLM, Kimi, GPT, Claude and others, in OpenAI and Anthropic formats. Cheaper tiers trade off stability; see the pricing page. |
| XHuoAPI (Standard) | deepseek-v4-flash | $0.24 / $0.96 | Saver $0.15 / $0.60 Spot $0.090 / $0.36 Official tier $0.30 / $1.20 The same key also calls Qwen, GLM, Kimi, GPT, Claude and others, in OpenAI and Anthropic formats. Cheaper tiers trade off stability; see the pricing page. |
DeepSeek V4 Pro
| Where | Model ID | $ / 1M in / out | Notes |
|---|---|---|---|
| DeepSeek (official) source | deepseek-v4-pro | $1.32 / $3.96 | Peak price. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday; all other hours (including weekends) are half price: $0.66 / $1.98 |
| DeepInfra (via OpenRouter) source | deepseek/deepseek-v4-pro | $1.30 / $2.60 | fp8 precision; price from DeepInfra's endpoint listed on OpenRouter. |
| OpenRouter source | deepseek/deepseek-v4-pro | — | This model has 15 providers on OpenRouter; the lowest in / out price is $0.90 / $1.50, and you pay the rate of the provider that serves the request. |
| XHuoAPI (Standard) | deepseek-v4-pro | $1.06 / $3.17 | Saver $0.66 / $1.98 Spot $0.40 / $1.19 Official tier $1.32 / $3.96 The same key also calls Qwen, GLM, Kimi, GPT, Claude and others, in OpenAI and Anthropic formats. Cheaper tiers trade off stability; see the pricing page. |
How to choose
- Lowest per-token price: in the table, OpenRouter's cheapest providers (mostly fp4 quantized) and XHuoAPI's Spot tier, followed by DeepInfra's promotional rate and DeepSeek's own off-peak rate. Spot is the cheapest tier and also the least stable.
- First-party, no intermediary: DeepSeek's platform, which offers both OpenAI and Anthropic formats.
- Switching between many vendors on one account: OpenRouter or XHuoAPI. On XHuoAPI, Flash in the Saver tier matches DeepSeek's off-peak rate at any hour, and one balance covers Chinese and global models.
- Coding tools such as Claude Code and OpenCode: both DeepSeek and XHuoAPI offer an Anthropic-format endpoint. Our test results for deepseek-v4.1-flash in both tools are on the test-results page linked below.
What to check besides price
- Precision: third-party hosts often serve fp8 or fp4 quantized versions, so output can differ slightly from the original.
- Time of day: DeepSeek's peak and off-peak prices differ by 2x; batch jobs are cheaper off-peak.
- Where data goes: a host or gateway adds another data processor. XHuoAPI runs in the US behind Cloudflare, and requests to Chinese vendors' models may be processed in mainland China; see the About page.