← TPUs, AI chips & CPU endpoints
Compare 4 tpus, ai chips & cpu endpoints APIs
| Cloud TPU v6e (Trillium) Google (Gemini / Vertex) | HF Endpoints CPU (4 vCPU) Hugging Face Inference Endpoints | HF Endpoints AWS Inferentia2 Hugging Face Inference Endpoints | HF Endpoints Google TPU Hugging Face Inference Endpoints | |
|---|---|---|---|---|
| $ / accel-hr | $2.70 Unverified sample | $0.134 Verified Oct 6, 2026 | $0.75 Verified Oct 6, 2026 | $1.20 Verified Oct 6, 2026 |
| Quality | – | – | – | – |
| Speed | – | – | – | – |
| Accelerator type | tpu | cpu | inferentia | tpu |
| Memory | 32 GB | 8 GB | — | — |
| Frameworks | jax, pytorch, tensorflow, vllm | onnx, pytorch | pytorch | jax, pytorch |
| Supports training | Yes | No | No | No |
| Per-minute billing | No | Yes | Yes | Yes |