Kimi K2.7 Code vs Kimi K3

One line per model: the best value available from any of its providers at each point, which is what a caller routing to the healthy endpoint would see.

CompareKimi K2.7 CodeKimi K3


Kimi K2.7 Code	1/21	100% Inceptron	99.87% Inceptron	99.59% Inceptron	801ms Inceptron	11.0s Inceptron	67 tps Inceptron	262K	$0.671 / $3.35 Inceptron
Kimi K3	14/29	100% DeepInfra	99.79% Wafer	99.79% Fireworks	606ms InferenceNet	2.64s Fireworks	95 tps Amazon Bedrock	1.0M	$0.657 / $13.00 Relace

Measured 1m ago. Each figure is the best available across that model's providers.

History

Availability

Best uptime available from any provider. Axis starts at 90%; it expands to 0 when a model falls below.

Time to first token

Fastest P50 available from any provider. Logarithmic axis.

Throughput

Best median tokens per second available.