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 |
z-ai
|
|---|---|
| Company | z-ai |
| Release Date | 2026-09-18 |
| Parameters (B) | 320 |
| Architecture | MoE Transformer with hybrid sparse and linear attention |
| Context Window | 1000000 |
| Input Modalities | text, image |
| Open Source | No |
| License | |
| Score | 42 |
| VRAM (GB) | — |
| Compute | API only |
| Benchmark: AIME-2026 | 92.7 |
| Benchmark: Agents-Last-Exam | 26.3 |
| Benchmark: Aider | 81.6 |
| Benchmark: Arena-Elo | 1541 |
| Benchmark: AutomationBench | 48.8 |
| Benchmark: BBH | 86.6 |
| Benchmark: BrowseComp | 75.9 |
| Benchmark: Context-Arena | 79.4 |
| Benchmark: DeepSWE | 63.4 |
| Benchmark: GDPval-AA | 1773 |
| Benchmark: GPQA-Diamond | 90.1 |
| Benchmark: HLE | 55.3 |
| Benchmark: HellaSwag | 87.1 |
| Benchmark: IFEval | 72 |
| Benchmark: IMO-AnswerBench | 82.5 |
| Benchmark: LiveBench | 78.9 |
| Benchmark: LiveCodeBench | 37.6 |
| Benchmark: MMLU | 88.1 |
| Benchmark: MMLU-Pro | 84.7 |
| Benchmark: NL2Repo-Bench | 56.3 |
| Benchmark: SWE-Bench-Verified | 77.8 |
| Benchmark: SimpleQA | 33.5 |
| Benchmark: Terminal-Bench-2.0 | 61.1 |
| Benchmark: Terminal-Bench-2.1 | 84.3 |
| Benchmark: Toolathlon-Verified | 78.4 |
| Pricing (Input) | 0.37 |
| Pricing (Output) | 1.25 |