DeepSeek V3 API pricing
Standard tier, blended at 3:1 input to output: $0.477 per million tokens.
DeepSeek charges $0.270 per million input tokens and $1.10 per million output tokens for DeepSeek V3. It is sold in 5 forms, from $0.477 to $1.25 blended; the table below shows what each costs against standard. 7 of the models nearest it in price are cheaper on a blended basis.
Every tier of DeepSeek V3
Blended at 3:1 input to output. The last column compares each against the standard tier.
| Source | ||||||||
|---|---|---|---|---|---|---|---|---|
| Standard Standard | deepseek-v3 | DeepSeek | $0.270 | $1.10 | $0.070 | $0.477 | Standard | ↗ |
| Standard | deepseek-v3 | Fireworks AI | $0.900 | $0.900 | Not listed | $0.900 | +88% | ↗ |
| Standard | DeepSeek-V3 | Together AI | $1.25 | $1.25 | Not listed | $1.25 | 2.6× | ↗ |
| Dated snapshot · Alias | deepseek-chat-v3-0324 | OpenRouter | $0.290 | $1.14 | $0.110 | $0.502 | +5% | ↗ |
| Dated snapshot | deepseek-v3-0324 | Fireworks AI | $0.900 | $0.900 | Not listed | $0.900 | +88% | ↗ |
Prices last checked: . Methodology
Which tier to use
What each costs against standard in our data, and the job it is for.
Dated snapshots
2 dated snapshots are listed at prices from 5% more than standard to 88% more than standard. A snapshot pins the model to one release, while the undated name moves to newer releases when the vendor updates it. Pin a snapshot in production when your prompts depend on exact output; use the undated name to pick up improvements automatically.
Aliases
"-latest" names point at whichever release the vendor currently serves, listed at 5% more than standard. The price can change under you when the alias moves, so budget against a pinned name.
Against the models nearest in price
Standard tiers, blended at a 3:1 input to output ratio.
| Source | ||||||
|---|---|---|---|---|---|---|
| DeepSeek V3 This model | DeepSeek | $0.270 | $1.10 | 65,536 | $0.477 | ↗ |
| GPT-5.4 nano | OpenAI | $0.200 | $1.25 | 272,000 | $0.463 | ↗ |
| GPT-5.6 Luna | OpenAI | $0.200 | $1.20 | 922,000 | $0.450 | ↗ |
| Llama 3.1 70B Instruct | OpenRouter | $0.400 | $0.400 | 131,072 | $0.400 | ↗ |
| Command A Plus | OpenRouter | $0.300 | $1.50 | 192,000 | $0.600 | ↗ |
| DeepSeek Chat | DeepSeek | $0.280 | $0.420 | 131,072 | $0.315 | ↗ |
| DeepSeek Reasoner | DeepSeek | $0.280 | $0.420 | 131,072 | $0.315 | ↗ |
| DeepSeek V3.2 | DeepSeek | $0.280 | $0.400 | 163,840 | $0.310 | ↗ |
| Llama 4 Maverick | OpenRouter | $0.188 | $0.652 | 524,288 | $0.304 | ↗ |
Prices last checked: . Methodology
Questions
How much does DeepSeek V3 cost per million tokens?
$0.270 per million input tokens and $1.10 per million output tokens, with cached input at $0.070. At a 3:1 input to output mix that blends to $0.477. Checked 28 Oct 2026.
What would DeepSeek V3 cost for a real workload?
A thousand requests a day at 2,000 input and 600 output tokens each comes to $36.00 a month on the standard tier.
What is the context window for DeepSeek V3?
65,536 tokens of input, with up to 8,192 tokens of output.
GPT-5.4 nano · GPT-5.6 Luna · Llama 3.1 70B Instruct · Command A Plus · DeepSeek Chat · All DeepSeek models · Cost calculator