Models / pythia
pythia by EleutherAI
19 models (14M to 12B) · ↓ 6.3M 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.
pythia 14M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-14m | vLLM | 30 MB | 2.3 GB | $0.10/h RTX 3060 |
pythia 14M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-14m-deduped | vLLM | 50 MB | 2.3 GB | $0.10/h RTX 3060 |
pythia 31M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-31m | vLLM | 60 MB | 2.3 GB | $0.10/h RTX 3060 |
pythia 70M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-70m-deduped | vLLM | 170 MB | 2.4 GB | $0.10/h RTX 3060 |
pythia 160M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-160m | vLLM | 370 MB | 2.7 GB | $0.10/h RTX 3060 |
pythia 160M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-160m-deduped | vLLM | 370 MB | 2.7 GB | $0.10/h RTX 3060 |
pythia 160M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-160m-seed1 | vLLM | 370 MB | 2.7 GB | $0.10/h RTX 3060 |
pythia 160M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-160m-seed2 | vLLM | 370 MB | 2.7 GB | $0.10/h RTX 3060 |
pythia 160M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-160m-seed3 | vLLM | 370 MB | 2.7 GB | $0.10/h RTX 3060 |
pythia 410M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-410m | vLLM | 910 MB | 3.5 GB | $0.10/h RTX 3060 |
pythia 410M
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-410m-deduped | vLLM | 910 MB | 3.5 GB | $0.10/h RTX 3060 |
pythia 1B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-1b | vLLM | 2.1 GB | 4.8 GB | $0.10/h RTX 3060 |
pythia 1B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-1b-deduped | vLLM | 2.1 GB | 4.8 GB | $0.10/h RTX 3060 |
pythia 1.4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-1.4b | vLLM | 2.9 GB | 5.9 GB | $0.10/h RTX 3060 |
pythia 1.4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-1.4b-deduped | vLLM | 2.8 GB | 5.8 GB | $0.10/h RTX 3060 |
pythia 2.8B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-2.8b | vLLM | 5.7 GB | 9.3 GB | $0.10/h RTX 3060 |
pythia 2.8B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-2.8b-deduped | vLLM | 5.7 GB | 9.3 GB | $0.10/h RTX 3060 |
pythia 6.9B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-6.9b | vLLM | 14 GB | 18.8 GB | $0.18/h 2x RTX 3060 |
pythia 12B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalEleutherAI/pythia-12b | vLLM | 24 GB | 30.4 GB | $0.44/h RTX 5000 Ada |
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").