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 |
moonshotai
|
|---|---|
| Company | moonshotai |
| Release Date | 2025-10-30 |
| Parameters (B) | 48 |
| Architecture | Hybrid KDA-MLA MoE |
| Context Window | 1000000 |
| Input Modalities | text |
| Open Source | Yes |
| License | mit |
| Score | 0 |
| VRAM (GB) | 96 |
| Compute | BF16 weights (~96 GB); recommended 4x NVIDIA GPUs with tensor parallelism (e.g., 4x A100 80GB) for vLLM deployment at 1M context |
| Benchmark: ARC | 67.3 |
| Benchmark: BBH | 72.9 |
| Benchmark: C-Eval | 79.5 |
| Benchmark: GSM8K | 83.9 |
| Benchmark: HellaSwag | 82.9 |
| Benchmark: MATH | 54.7 |
| Benchmark: MMLU | 73.8 |
| Benchmark: MMLU-Pro | 51 |
| Pricing (Input) | — |
| Pricing (Output) | — |