DeepSeek's DeepSeek and OpenAI's GPT, compared through their current models on the Respan Index. The strongest DeepSeek model today is DeepSeek V4.1 Flash (#20 on the Respan Index, 50.9); the strongest GPT model is GPT-6 Astra (#2, 70.5). GPT's lead in coding, agents, reasoning, knowledge and human preference.
| Category | DeepSeek | GPT |
|---|---|---|
| coding | DeepSeek V4.1 Flash 40.1 | GPT-6 Astra 61.8 |
| agents | DeepSeek V4.1 Flash 61.2 | GPT-6 Astra 77.0 |
| reasoning | DeepSeek V4.1 Flash 48.2 | GPT-6.1 Sol 78.6 |
| knowledge | DeepSeek V4 Pro 52.6 | GPT-6 Astra 75.8 |
| human preference | DeepSeek V4.1 Flash 55.7 | GPT-6 Astra |
Each family's highest Respan Index category score; the higher one is bold. * Estimated: no published scores in that category. How the Respan Index works
Respan Index rank and score, blended list price per 1M tokens.
Respan Index rank and score, blended list price per 1M tokens.
The cheapest ranked DeepSeek model is DeepSeek V4.1 Flash at $0.525 per 1M tokens blended; the cheapest GPT model is GPT-5 nano at $0.138.