Inkling (Thinking Machines) is a 975.0B parameters MoE (41.0B active per token) open-weight model tracked on AI Hashrate. At 8K context we estimate roughly 545.45 GB VRAM for Q4 (weights 536.25 GB + KV 8.2 GB + 1 GB overhead) and 1959.2 GB for FP16. About 0 GPUs/accelerators in our catalog fully fit this model at Q4 under that context assumption. Fastest Q4 configs: B200 192GB SXM ≈ 44.0 tok/s (estimated); MI300X 192GB ≈ 29.1 tok/s (estimated); H200 141GB SXM5 ≈ 14.2 tok/s (estimated). Use the table below for VRAM fit and tok/s per dollar (list MSRP). Estimates follow memory-bandwidth math; measured rows override when present. See methodology for details. Methodology.
| Hardware | VRAM | Quant | tok/s | Fits? | tok/s/$ |
|---|---|---|---|---|---|
| B200 192GB SXM | 192 | Q4 | 44.0 est. | No | 0.001 |
| MI300X 192GB | 192 | Q4 | 29.1 est. | No | 0.002 |
| H200 141GB SXM5 | 141 | Q4 | 14.2 est. | No | 0.0 |
| Mac Studio M2 Ultra 192GB | 192 | Q4 | 4.4 est. | No | 0.001 |