Compare LLM Models
Select 2-5 models to compare side by side.
Select models to compare
0 / 5 selected
Top downloads
Select 2-5 models
| Attribute |
deepseek-ai
|
|---|---|
| Company | deepseek-ai |
| Release Date | 2023-11-29 |
| Parameters (B) | 67 |
| Architecture | Transformer |
| Context Window | 4096 |
| Input Modalities | text |
| Open Source | Yes |
| License | other |
| Score | 0 |
| VRAM (GB) | 134 |
| Compute | Approximately 134 GB VRAM at BF16 precision for full-weight inference; typically requires multi-GPU setup (e.g., 2x NVIDIA A100 80GB or H100). INT4 quantization reduces requirements to roughly 34 GB VRAM. |
| Benchmark: AGIEval | 41.3 |
| Benchmark: ARC | 59 |
| Benchmark: BBH | 68.7 |
| Benchmark: C-Eval | 66.1 |
| Benchmark: DROP | 67.9 |
| Benchmark: GSM8K | 63.4 |
| Benchmark: HellaSwag | 84 |
| Benchmark: HumanEval | 42.7 |
| Benchmark: MATH | 18.7 |
| Benchmark: MBPP | 57.4 |
| Benchmark: MMLU | 71.3 |
| Pricing (Input) | — |
| Pricing (Output) | — |