Act on this, mostly signal

Kimi K3.1 Pro and Highspeed roll out in Kimi Code

What to do

Try K3.1 Pro in Kimi Code if you need a 1M context model, or Highspeed for about 2 to 3 times the speed.

Why

Kimi Code now serves a 3T-class model with 1M context and a thinking-only mode, plus a faster variant with 2 times the quota.

Read the source at kimi.com Discussion

What the research found

Kimi Code docs show K3.1 Pro rolled out as a 3T-class model with 1M context and thinking-only mode, plus K3.1 Pro Highspeed at about 2-3x speed and 2x quota; same model as #5732 so the two are one story.

Claims and how well the evidence supports them

K3.1 Pro has a 1M context window and thinking mode only 99%
K3.1 Pro Highspeed runs about 2-3x faster with about 2x quota usage 99%
Requests with thinking disabled are rejected for K3.1 Pro models 99%

Evidence

kimi.com https://kimi.com/code/docs/en/kimi-code/models.html
news.ycombinator.com https://news.ycombinator.com/item?id=50212065

Actionability rated high.