Kimi K2.6: Fast - Cogito

The fast route for Kimi K2.6 runs the same NVIDIA NVFP4 target as Kimi K2.6 on a dedicated AWS B300 SGLang lane. One EP8 prefill worker and one EP8 decode worker isolate its low-latency capacity from the default route.


Model ID	`moonshotai/kimi-k2.6:fast`
Route slug	`kimi-k2.6-fast`
Parameters	1T MoE (~32B active)
Context	262,144 tokens (256K)
Hardware	AWS B300
Serving topology	1× EP8 prefill + 1× EP8 decode
License	Modified MIT
Pricing	cogito.decart.ai/models/kimi-k2.6-fast

Best for

Latency-sensitive coding and tool-use agents
Workloads that benefit from isolated route capacity
The same target model as Kimi K2.6 on a dedicated serving lane

client.chat.completions.create(
    model="moonshotai/kimi-k2.6:fast",
    messages=[...],
)

License

Same as Kimi K2.6 — Modified MIT.