Model reference · open weights
HarmBench-Llama-2-cls is an open-weight language model from cais. HarmBench-Llama-2-13b-cls (BF16) weighs 26.0 GB; the smallest configuration that runs it is 2× RTX 4060 Ti 16 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the cais/HarmBench-Llama-2-13b-cls model card, 2026-10-01
What it is
| Released by | cais |
|---|---|
| Released | 2024-02-03 |
| Parameters | 13.0B |
| VRAM | 26.0 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory |
|---|---|---|---|
| 2K each | |||
| RTX 3060 12 GB … RTX 4090 24 GB 4 smaller cards | |||
| RTX 5090 32 GB | 2 | all 2K | 31.0 GB |
| L40S 48 GB | 10 | all 2K | 44.0 GB |
| A100 80 GB | 30 | all 2K | 78.2 GB |
| H100 80 GB | 28 | all 2K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 38 | all 2K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 46 | all 2K | 107 GB |
| H200 141 GB | 64 | all 2K | 138 GB |
| B200 180 GB | 86 | all 2K | 176 GB |
| 2× RTX 4060 Ti 16 GB tensor parallel | 2 | all 2K | 15.4 GB a card |
| 2× RTX 4090 24 GB tensor parallel | 11 | all 2K | 23.4 GB a card |
| 2× RTX 3090 24 GB tensor parallel | 11 | all 2K | 23.4 GB a card |
| 2× RTX 5090 32 GB tensor parallel | 20 | all 2K | 31.0 GB a card |
| Requests at once | 2K tokens each |
|---|---|
| 1 | 28.4 GB |
| 5 | 35.1 GB |
| 8 | 40.1 GB |
| 16 | 53.5 GB |
| 32 | 80.4 GB |
| 64 | 134 GB |
One card, with vLLM's small-card settings.