DeepSeek V3.2 vs DeepSeek V4 Pro
deepseek/deepseek-v3.2 · deepseek/deepseek-v4-pro
As of November 20, 2026, DeepSeek V3.2 is the cheaper API on MartLLM: $0.0026 per million input tokens and $0.0038 per million output, versus $0.0634 / $0.1268 for DeepSeek V4 Pro, about 97% less when output is 3 tokens for every input token. These are the best prices on the market right now. Both work with one OpenAI-compatible key and MartLLM adds no fee. Choose DeepSeek V3.2 to cut cost; choose DeepSeek V4 Pro when context size wins. DeepSeek V4 Pro carries the larger context window (1,000,000 vs 160,000 tokens), which can matter for long-document work.
DeepSeek V3.2 vs DeepSeek V4 Pro, side by side
| Spec | ||
|---|---|---|
| Price · input / 1M | $0.0026 | $0.0634 |
| Price · output / 1M | $0.0038 | $0.1268 |
| Context length | 160,000 | 1,000,000 |
| Max output tokens | 32,768 | 128,000 |
| Input modalities | text | text |
| Output modalities | text | text |
| Tools / functions | Yes | Yes |
| Reasoning | Yes | Yes |
| % off list | 99% off | 95.1% off |
| First token, median (7 days) | 5.4 s (3,363 requests) | 6.7 s (4,829 requests) |
Best prices on MartLLM right now. No MartLLM fee on requests; adding credit by card costs 5.5% + $0.50. Updated November 20, 2026. A price more than 70% below list is what buyers were billed, not an error. Why so cheap?
DeepSeek V3.2 vs DeepSeek V4 Pro — FAQ
Is DeepSeek V3.2 or DeepSeek V4 Pro cheaper?
DeepSeek V3.2 is cheaper on a blended token mix, by about 97%. A request costs what its provider charged, MartLLM adds no fee, and every request is listed with its cost in Activity.
What's the price difference between DeepSeek V3.2 and DeepSeek V4 Pro?
DeepSeek V3.2: $0.0026 in / $0.0038 out per 1M tokens. DeepSeek V4 Pro: $0.0634 in / $0.1268 out per 1M tokens. That's roughly a 97% blended difference. Prices are live; MartLLM adds no fee.
Can I use both DeepSeek V3.2 and DeepSeek V4 Pro with one API?
Yes. Both are served through a single OpenAI-compatible endpoint — set the base URL to https://api.martllm.com/v1 and your MartLLM key, then switch between them by changing the model id. Anthropic SDKs work via https://api.martllm.com/v1/messages.