Fireworks AI · none caching
KAT Dev 72B Exp
KAT Dev 72B Exp 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 | qwen2 |
| Released | — |
| License | Open weights |
| Context window | 131K tokens |
| Max output | — |
| Parameters | 73B |
| Modalities | text |
| Tool calling | Yes |
| Reasoning mode | No |
| Caching | none |
| Batch discount | No batch tier |
Measured by Zumik
What reuse looks like here.
Pricing, context, and capabilities for KAT Dev 72B Exp 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.
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="kat-dev-72b-exp",
input="Draft a fix for the failing test.",
)
print(r.usage.input_tokens_cached) # confirm reuseFrequently asked
KAT Dev 72B Exp, answered.
How much does KAT Dev 72B Exp cost?
KAT Dev 72B Exp is an open-weights model routed through Fireworks AI. It is priced on the host's serverless size tier rather than a single published per-token list price, so it shows "—" here until profiled.
What is KAT Dev 72B Exp's context window?
KAT Dev 72B Exp supports a 131K-token context window.
Does KAT Dev 72B Exp support prompt caching?
Yes. Fireworks AI uses Automatic prompt caching (serverless and dedicated) caching. In the Zumik corpus, KAT Dev 72B Exp shows a median cache capture of 78% on agent workloads.
Run KAT Dev 72B Exp with reuse measured.
Point an OpenAI client at Zumik and see exactly how much of this model's input you are reusing.