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 |
Qwen
|
|---|---|
| Company | Qwen |
| Release Date | 2026-02-03 |
| Parameters (B) | 80 |
| Architecture | Hybrid MoE Transformer |
| Context Window | 262144 |
| Input Modalities | text |
| Open Source | Yes |
| License | apache-2.0 |
| Score | 21 |
| VRAM (GB) | 80 |
| Compute | 80B sparse MoE (3B active/token), fine-grained FP8 weights (~80 GB); recommended 2-GPU tensor parallel (e.g. 2× A100/H100 80GB) for 256K-context serving via vLLM or SGLang |
| Benchmark: AIME-2024 | 89.0 |
| Benchmark: AIME-2025 | 83.1 |
| Benchmark: Aider | 66.2 |
| Benchmark: GPQA | 74.5 |
| Benchmark: LiveCodeBench | 58.9 |
| Benchmark: MMLU | 87.7 |
| Benchmark: MMLU-Pro | 80.5 |
| Benchmark: SWE-Bench-Pro | 42.7 |
| Benchmark: SWE-Bench-Verified | 70.6 |
| Pricing (Input) | — |
| Pricing (Output) | — |