DeepSeek V3 was retired from the DeepSeek API on July 24, 2026. Prices shown are the last published rates. DeepSeek names DeepSeek V4.1 Flash as the replacement. Details
Cost Comparison
| Prompts | DeepSeek V4.1 Flash | DeepSeek V3 |
|---|---|---|
| 1 | $0.000750 | $0.000350 |
| 100 | $0.0750 | $0.0350 |
| 1,000 | $0.7500 | $0.3500 |
| 10,000 | $7.50 | $3.50 |
| 100,000 | $75.00 | $35.00 |
Based on 500 input + 500 output tokens per prompt. Prices verified October 2026. Embed this calculator
At 10,000 requests per day, DeepSeek V4.1 Flash costs $225.00/month vs DeepSeek V3 at $105.00/month — DeepSeek V3 is 53% cheaper.
| DeepSeek V4.1 Flash | DeepSeek V3 | |
|---|---|---|
| Input price | $0.30 / 1M | $0.28 / 1M |
| Output price | $1.20 / 1M | $0.42 / 1M |
| Cached input | $0.006 / 1M | — |
| Typical prompt | $0.000750 | $0.000350 |
| Batch discount | None | None |
| Context window | 1M tokens | 131K tokens |
| Max output | 384,000 tokens | Not published |
| Knowledge cutoff | Not published | Not published |
| Input types | Text, Images | Text |
| Reasoning mode | Optional | No |
| Open weights | Yes | Yes |
| Released | September 10, 2026 | December 1, 2025 |
| Status | Current | Retired |
Try It Yourself
How many tokens you expect the model to generate per response
Showing a typical prompt (500 input + 500 output tokens) across 27 models. Paste your own prompt above for an exact count.
| Model | Per Prompt | Prompts / $1 | Monthly |
|---|---|---|---|
DeepSeek V3Retired DeepSeek | $0.0003500.035¢ | 2,857 | $1.05 |
| DeepSeek | $0.0007500.075¢ | 1,333 | $2.25 |
GPT-6 LunaCheapest OpenAI | $0.0003000.03¢ | 3,333 | $0.9000 |
| Mistral | $0.0003750.037¢ | 2,666 | $1.13 |
| OpenAI (via Together) | $0.0003750.037¢ | 2,666 | $1.13 |
| OpenAI | $0.0007000.07¢ | 1,428 | $2.10 |
| Meta (via Together) | $0.0009250.092¢ | 1,081 | $2.77 |
| Mistral | $0.00100.1¢ | 1,000 | $3.00 |
| $0.00140.14¢ | 714 | $4.20 | |
| xAI | $0.00190.19¢ | 533 | $5.63 |
| $0.00220.22¢ | 444 | $6.75 | |
| DeepSeek | $0.00260.26¢ | 378 | $7.92 |
| Meta | $0.00270.27¢ | 363 | $8.25 |
| Anthropic | $0.00300.3¢ | 333 | $9.00 |
| xAI | $0.00400.4¢ | 250 | $12.00 |
| Alibaba (Qwen) | $0.00400.4¢ | 250 | $12.00 |
| Mistral | $0.00450.45¢ | 222 | $13.50 |
| OpenAI | $0.00600.6¢ | 166 | $18.00 |
| OpenAI | $0.00600.6¢ | 166 | $18.00 |
| Anthropic | $0.00600.6¢ | 166 | $18.00 |
| OpenAI | $0.00700.7¢ | 142 | $21.00 |
| $0.00700.7¢ | 142 | $21.00 | |
| Moonshot AI | $0.00900.9¢ | 111 | $27.00 |
| OpenAI | $0.0120 | 83 | $36.00 |
| Anthropic | $0.0120 | 83 | $36.00 |
| OpenAI | $0.0300 | 33 | $90.00 |
| Anthropic | $0.0300 | 33 | $90.00 |
Prices from official provider pricing pages, verified October 2026.
DeepSeek V4.1 Flash vs DeepSeek V3 — Frequently Asked Questions
DeepSeek V3 is 53% cheaper than DeepSeek V4.1 Flash. One typical prompt (500 input + 500 output tokens) costs $0.000750 on DeepSeek V4.1 Flash and $0.000350 on DeepSeek V3. DeepSeek V4.1 Flash charges $0.30/$1.20 per million tokens (input/output) vs DeepSeek V3's $0.28/$0.42.
It depends on your priorities. DeepSeek V3 is more cost-effective at $0.3500 per 1,000 prompts. DeepSeek V4.1 Flash may offer different capabilities or quality. DeepSeek V4.1 Flash has a 1M context window vs DeepSeek V3's 131K. Use the calculator above to compare at your actual volume.
Beyond price: DeepSeek V4.1 Flash accepts text and images, writes up to 384,000 tokens per response, has an optional reasoning (thinking) mode and has open weights you can self-host. DeepSeek V3 accepts text, has no reasoning mode and has open weights you can self-host. DeepSeek V4.1 Flash has a 1M context window and DeepSeek V3 131K. See the specs table above for a side-by-side view.
At 10,000 requests per day, switching from DeepSeek V4.1 Flash to DeepSeek V3 saves about $120.00/month — a 53% cut on every call. Savings scale linearly with volume.
At 100,000 daily requests, DeepSeek V4.1 Flash costs $2,250.00/month and DeepSeek V3 costs $1,050.00/month — a difference of $1,200.00/month.