Fireworks AI · automatic caching
Kimi K3
Kimi K3 on Zumik: live pricing, context, and caching, routable by id or alias through one OpenAI-compatible endpoint.
Specifications
At a glance.
| Provider | Fireworks AI |
| Family | kimi_k3 |
| Released | 2026-07 |
| License | Open weights |
| Context window | 1M tokens |
| Max output | 131K tokens |
| Parameters | 2812B |
| Modalities | text, image |
| Tool calling | Yes |
| Reasoning mode | Yes |
| Caching | automatic |
| Batch discount | No batch tier |
Measured by Zumik
What reuse looks like here.
Pricing, context, and capabilities for Kimi K3 are live, but it is outside the flagship set Zumik benchmarks in depth, so measured reuse, capture, and warm TTFT are not shown yet. Run a workload estimate or route it by id to start collecting traces.
What you actually pay once caching works.
At a typical 55% prefix reuse, a million input tokens on Kimi K3 effectively costs $1.51 instead of $3.00 - blending to roughly $4.89 with a 25% output share. There is no batch tier, so cost control here leans on caching and routing.
Estimate it for your workloadRoute it directly by id, or let an alias pick it when it wins under policy.
Call it
Same OpenAI client, this model.
from openai import OpenAI
client = OpenAI(base_url="https://api.zumik.ai/v1", api_key="zk_live_...")
r = client.responses.create(
model="kimi-k3",
input="Draft a fix for the failing test.",
)
print(r.usage.input_tokens_cached) # confirm reuseFrequently asked
Kimi K3, answered.
How much does Kimi K3 cost?
Kimi K3 is $3.00 per million input tokens and $15.00 per million output tokens through Zumik. Cache reads are $0.30 per million, a 90% discount on input.
What is Kimi K3's context window?
Kimi K3 supports a 1M-token context window with up to 131K output tokens.
Does Kimi K3 support prompt caching?
Yes. Fireworks AI uses Automatic prompt caching (serverless and dedicated) caching. In the Zumik corpus, Kimi K3 shows a median cache capture of 84% on agent workloads.
Run Kimi K3 with reuse measured.
Point an OpenAI client at Zumik and see exactly how much of this model's input you are reusing.