Kimi K3.1 Pro Highspeed

PRICE / USD

Not verified

Per 1 million uncached input tokens · output billed separately

Input / 1M tokens

Not verified

Output / 1M tokens

Not verified

Cached input / 1M tokens

Not verified

Cache write / 1M tokens

Not verified

The launch post prices the Highspeed variant at twice the standard K3.1 Pro rates — about ¥2.6 cache-hit input, ¥26 cache-miss input and ¥136 output per million tokens — without showing a separate listing.

SPEED & CAPABILITIES

Not benchmarked

Moonshot states roughly two to three times the standard model's output speed. No absolute tokens-per-second figure was published, so no verified speed classification exists.

Our own output-speed measurement

Not measured

Context window

1,048,576

API model ID

kimi-k3.1-pro-highspeed

No reviewed Intelligence Index score for this exact model version and setting.

Shares the K3.1 Pro post-training; only serving speed and price differ, per the launch post.

Suggested starting role: everyday coding. The Highspeed serving option targets interactive coding and agent loops where output latency feeds back into cost or waiting time; offline batch work can stay on the standard model at half the price. This is editorial guidance, not a benchmark score.