Not measured here yet. OpenRouter sells it at $0.14 per million output tokens.
2 builds of Phi 4 BASE, 2 signed runs on real machines: different packagers, precisions and engines, each a different file with its own numbers.
proved 32K context, by redhatai
What it fits on
The smallest build needs about 32.8 GB; 1 of 18 kinds of machine on the network can hold it. Bars: each kind's memory (its card's, else RAM) · ticks: every build, at the size it held plus a fifth for the prompt.
How far back it remembers, machine by machine
Filled: the needle came back at that prompt size · outlined: asked and lost · blank: not asked. 0 walks ended because the model forgot, 1 because the machine could not hold the prompt.
| machine · build | 4K | 8K | 16K | 32K | 64K | why it stopped |
|---|---|---|---|---|---|---|
| Gram ·a5be462× Tesla T4 · by redhatai | 32K this machine could not serve at 64K |
Measured by
The people whose machines signed these numbers
proved by at least one build: Structured output · Sustained output · Writes code · Reasons about code · Synthesises sources · Follows instructions
Each its own file with its own page: the machines that ran it and their signatures.
| build | engine | whose work | memory | context proved | best tok/s | machines |
|---|---|---|---|---|---|---|
| by redhataiRedHatAI/phi-4-quantized.w8a8 | vllm | redhatai | — | 32K | — | 1 |
| BF16 by microsofthf.co/microsoft/phi-4-gguf:BF16 | ollama | vendor · microsoft | 32.8 GB | — | 3.6 | 1 |