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 |
ibm-granite
|
|---|---|
| Company | ibm-granite |
| Release Date | 2025-10-02 |
| Parameters (B) | 3 |
| Architecture | Hybrid Mamba-2/Transformer |
| Context Window | 128000 |
| Input Modalities | text |
| Open Source | Yes |
| License | apache-2.0 |
| Score | 2 |
| VRAM (GB) | 8 |
| Compute | 3B-parameter dense hybrid model in BF16; runs on a single GPU with ~8 GB VRAM (or ~2–3 GB with Q4 quantization). CPU inference is supported but slower. Optimized for low-latency local and edge deployment. |
| Benchmark: AGIEval | 59 |
| Benchmark: BBH | 69.4 |
| Benchmark: GPQA | 32.1 |
| Benchmark: GSM8K | 81.3 |
| Benchmark: HumanEval | 81 |
| Benchmark: IFEval | 84.3 |
| Benchmark: MBPP | 73 |
| Benchmark: MMLU | 67.4 |
| Benchmark: MMLU-Pro | 43.5 |
| Pricing (Input) | 0.02 |
| Pricing (Output) | 0.11 |