Kimi K3

PRICE / USD

$3

Per 1 million output tokens · input billed separately

Input / 1M tokens

$3

Output / 1M tokens

$15

Cached input / 1M tokens

$0.3

Cache write / 1M tokens

$3

Cache write shown for 5-minute TTL; 1-hour writes cost $6/MTok. Tax is excluded.

SPEED & CAPABILITIES

Not benchmarked

No speed classification verified for this version and serving tier. This does not mean it is slow.

Model score

44

Our own output-speed measurement

Not measured

Context window

1,048,576

API model ID

kimi-k3

Artificial Analysis tested Kimi K3 · max. This mostly English, text-based index is not a guarantee of performance on your workload. Artificial Analysis · Kimi K3 · max ↗

K3 is a separate model from the K2.6 historical evaluation.

Suggested starting role: complex planning. The documented focus on long-horizon software engineering and reasoning makes K3 a planning and extended-agent candidate. Its large context is not itself evidence of higher accuracy. This is editorial guidance, not a benchmark score.