Compare, 2 models
Side by side.
DeepSeek-V4.1-Flash and GPT-5.2 on every board at least two of them share. Blue marks the best in each row.
Specs
List prices, context and published output speed. Blue marks the best in each row.
| Spec | DeepSeek-V4.1-FlashDeepSeek | GPT-5.2OpenAI |
|---|---|---|
| Lab | DeepSeek | OpenAI |
| Type | Language | Language |
| Released | Sep 10, 2026 | Dec 11, 2025 |
| Input price$ per 1M tokens | $0.30 | $1.75 |
| Output price$ per 1M tokens | $1.20 | $14.00 |
| Blended price3 input tokens per output | $0.52 | $4.81 |
| Contexttokens | 1.05M | 400K |
| Open weights | Yes | No |
| Output speedTok / s | 202 | n/a |
Benchmarks
Every board where at least two of these models have a published score. Each score links to its source.
| Board | DeepSeek-V4.1-FlashDeepSeek | GPT-5.2OpenAI |
|---|---|---|
| Epoch Capabilities IndexOverall, higher is better | 156 | 154 |
| ARC-AGI-2Reasoning and knowledge, higher is better | 72.9%max setting | 52.9%xhigh setting |
| FrontierMath Tiers 1-3Reasoning and knowledge, higher is better | 67.4%max setting | 67.4%xhigh setting |
| GPQA DiamondReasoning and knowledge, higher is better | 89.8%max setting | 91.4%xhigh setting |
n/a means the board has not published a score for that model.