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 |
meituan-longcat
|
|---|---|
| Company | meituan-longcat |
| Release Date | 2026-01-29 |
| Parameters (B) | 68.5 |
| Architecture | MoE Transformer with N-gram Embedding |
| Context Window | 256000 |
| Input Modalities | text |
| Open Source | Yes |
| License | mit |
| Score | 11 |
| VRAM (GB) | 160 |
| Compute | Minimum 2× NVIDIA H100 or A100 80GB GPUs for local inference via Transformers; SGLang deployment recommended on a single 8×H20 (141GB) node with tensor parallelism (TP=8) and expert parallelism (EP=8) |
| Benchmark: AIME-2024 | 70.5 |
| Benchmark: AIME-2025 | 59.6 |
| Benchmark: C-Eval | 78.5 |
| Benchmark: GPQA-Diamond | 69.9 |
| Benchmark: MATH-500 | 94.2 |
| Benchmark: MMLU | 79.9 |
| Benchmark: MMLU-Pro | 67.2 |
| Benchmark: SWE-Bench | 32.8 |
| Benchmark: Terminal-Bench-2.1 | 20 |
| Pricing (Input) | — |
| Pricing (Output) | — |