Kimi K3.1 Pro shifts the constraint to active parameters
The 104 billion active parameters, not the 2.8 trillion total, set the memory and compute pressure.
| Kimi K3.1 Pro | |
|---|---|
| Total parameters | 2.8 trillion |
| Active parameters | 104 billion per token |
| Context window | 1 million tokens |
| Input price | $2 per 1M tokens |
| Output price | $10 per 1M tokens |
| Cache-hit price | $0.20 per 1M tokens |
| Open weights date | November 04 |
| Intelligence Index | 55 |
The 1 million token native context and the 55 Artificial Analysis Intelligence Index score round out the parameter count.
The document prices it at $2, $10, and $0.20 per million input, output, and cache-hit tokens.
The open-weights date is November 04, but the document does not name a license.
MXFP4 weights and MXFP8 activations set the memory and compute pressure behind the 1.8 percent activation ratio.
The exact license, launch date, energy use, node count, SKU, and independent verification of choblin29's cyber-data stripping claim remain unmeasured.
Lead with active parameters and pricing, and flag the missing license as the buyer constraint.
After @ArtificialAnlys. We did not report this. The pictures, if any, are theirs.
