DeepSeek V3.2 Thinking vs DeepSeek V4.1 Flash 1017
Lower and Larger mark one number at a time; they don’t rank the models. How well a model does depends on your task; see Benchmarks for measured results.
| Detail | DeepSeek V3.2 Thinkingdeepseek/deepseek-v3.2-thinking | DeepSeek V4.1 Flash 1017deepseek/deepseek-v4.1-flash-1017 |
|---|---|---|
| Maker | deepseek | deepseek |
| Kind | Chat | Chat |
| Input per 1M tokens | $0.651 | $0.07875 Lower |
| Cached input per 1M tokens | Not applicable | $0.001575 |
| Cache write per 1M tokens | Not applicable | Same as input |
| Output per 1M tokens | $1.9425 | $0.315 Lower |
| Context window | 128,000 tokens | 1,048,576 tokens Larger |
| Max output | 8,000 tokens | 262,144 tokens Larger |
| Capabilities | Tool use | Reasoning, tool use, vision |
| Takes | Text | Text, image |
| Returns | Text | Text |
| Zero data retention | Yes, on every provider | Not on every provider |
| No training on requests | Yes, on every provider | Not on every provider |
| Released | 2025-12-01 | 2026-10-17 |
Example costs
Example requests, priced at today’s rates with no cached input.
| Example request | DeepSeek V3.2 Thinking | DeepSeek V4.1 Flash 1017 |
|---|---|---|
| 1,000 tokens in, 200 out | $0.00104 | $0.000142 |
| 10,000 tokens in, 1,000 out | $0.008453 | $0.001103 |
| 100,000 tokens in, 4,000 out | $0.07287 | $0.009135 |