Model vendor
Chinese AI company developing open and frontier language models, including the DeepSeek-V3 and DeepSeek-R1 model families.
Updated Aug 11, 2026. Data as of 2026-08-11
Default-version specifications and scored-version capability values are labeled separately.
| DE | DeepSeek-V4-Flash-0731 | 78.3 | 20.0% | 1M | Open weights | DeepSeek-V4-Flash-0731 | |
| DE | DeepSeek V4 Pro | 73.8 | 80.0% | 1M | Open weights | DeepSeek-V4-Pro-Max | |
| DE | DeepSeek-V3.2-Speciale | 58.1 | 60.0% | 131.1K | Closed weights | DeepSeek-V3.2-Speciale | |
| DE | DeepSeek-R1-0528 | 44.2 | 80.0% | 131.1K | Closed weights | DeepSeek-R1-0528 | |
| DE | DeepSeek-V3.1 | 39.9 | 80.0% | 131.1K | Open weights | DeepSeek-V3.1 | |
| DE | DeepSeek-V3 0324 | 26.0 | 80.0% | 163.8K | Closed weights | DeepSeek-V3 0324 | |
| DE | DeepSeek R1 Distill Qwen 7B | 15.7 | 60.0% | N/A | Closed weights | DeepSeek R1 Distill Qwen 7B | |
| DE | DeepSeek-V2.5 | 15.4 | 100.0% | 8.2K | Closed weights | DeepSeek-V2.5 | |
| DE | DeepSeek R1 Distill Llama 8B | 12.4 | 60.0% | N/A | Closed weights | DeepSeek R1 Distill Llama 8B | |
| DE | DeepSeek VL2 Tiny | 0.0 | 0.0% | N/A | Closed weights | DeepSeek VL2 Tiny |
Each unique model uses its latest scored version.
Leading DeepSeek models by score, price and context window.
Background and model catalog information for DeepSeek.
Chinese AI company developing open and frontier language models, including the DeepSeek-V3 and DeepSeek-R1 model families.
Model ownership is separate from API hosting. One model can be offered by multiple API providers with different prices.
Catalog data as of 2026-08-11.
Common questions about DeepSeek.
DeepSeek-V4-Flash currently has the strongest LLMBoard score from this vendor.
The catalog contains 10 independently selectable DeepSeek models and 26 versions.
Not necessarily. Vendor identifies the model developer; API providers are tracked separately.
3 current default versions are marked as open weight.