12models tracked · ranked by intelligence, speed & price
DeepSeek's smartest model is DeepSeek V4 Flash (46.0 on the Intelligence Index), its fastest is DeepSeek V3.1 at 85 tokens/sec, and its cheapest is DeepSeek V4 Flash at $0.098 per million input tokens. All 12 DeepSeek models are compared below by intelligence, speed, latency, context and price.
| Model | Intel | Speed | Latency | In $/M | Context |
|---|---|---|---|---|---|
| deepseek-v4-flash ReasoningToolsJSON | 46.0 | 79 | 582ms | $0.098 | 1.0M |
| deepseek-v3.2 ReasoningToolsJSON | 41.7 | 83 | 688ms | $0.229 | 131K |
| deepseek-v4-pro ReasoningToolsJSON | 39.3 | 62 | 604ms | $0.435 | 1.0M |
| deepseek-v3.2-exp ReasoningToolsJSON | 32.1 | 18 | 1.2s | $0.270 | 164K |
| deepseek-v3.1-terminus ReasoningToolsJSON | 28.5 | 26 | 1.1s | $0.270 | 164K |
| deepseek-chat-v3.1 ReasoningToolsJSON | 28.1 | 85 | 368ms | $0.210 | 164K |
| deepseek-r1-0528 ReasoningToolsJSON | 27.1 | 30 | 627ms | $0.500 | 164K |
| deepseek-chat-v3-0324 ToolsJSON | 22.3 | 35 | 900ms | $0.200 | 164K |
| deepseek-r1 ReasoningToolsJSON | 18.8 | 70 | 1.5s | $0.700 | 164K |
| deepseek-r1-distill-qwen-32b ReasoningJSON | — | 23 | 824ms | $0.290 | 128K |
| deepseek-r1-distill-llama-70b Reasoning | — | 26 | 870ms | $0.800 | 128K |
| deepseek-chat ToolsJSON | — | 24 | 854ms | $0.200 | 131K |