Loading…
Loading…
Models / starcoder2
3 models (3B to 15B) · ↓ 72.8K downloads a month · from $0.10/h on RTX 3060 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalbigcode/starcoder2-3b | vLLM | 6.1 GB | 9.2 GB | $0.10/h RTX 3060 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalbigcode/starcoder2-7b | vLLM | 14 GB | 18.8 GB | $0.16/h 2x RTX 3060 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalbigcode/starcoder2-15b | vLLM | 32 GB | 38.4 GB | $0.45/h 2x RTX 3090 |
| instruct v0.1bigcode/starcoder2-15b-instruct-v0.1 | vLLM | 32 GB | 38.4 GB | $0.45/h 2x RTX 3090 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").