Three model sizes, one API surface, and tooling that behaves the same way across all of them. The interesting part is supposed to be what you build, not which endpoint you called.
Pick by shape of the task, not by leaderboard position. All three share a context window, a tool protocol, and a refusal policy.
Multi-day agentic runs, ambiguous briefs, and problems where a wrong confident answer costs more than a slow careful one. Slower and more expensive, and worth it when it is.
Most production traffic should land here. Fast enough for interactive use, careful enough for customer-facing work, and the model we tune first when a new capability ships.
Classification, routing, and the million small calls that make an agent feel responsive. Cheap enough that you design around latency instead of budget.
Swap the model name and the rest of your code keeps working. Tool definitions, streaming, and structured output behave identically across the lineup, so a prototype can graduate without a rewrite.
from anthropic import Neurok
client = Neurok()
# Same call shape at every size — swap the model, keep the code.
response = client.messages.create(
model="neurok-pulse-5-1",
max_tokens=2048,
tools=[search_docs, run_query],
messages=[
{"role": "user",
"content": "Reconcile last quarter and flag anything odd."}
],
)
for block in response.content:
if block.type == "tool_use":
print("calling", block.name, block.input)Published rates, no sales call required for standard usage. Enterprise adds controls, not a different model.
| Plan | Price | Includes | Best for |
|---|---|---|---|
| Free | $0 | Pulse 5.1 with daily limits, web search, file uploads. | Trying it on real work |
| Pro | $20 / month | Cortex 5.1 access, higher limits, projects, shared context, desktop app. | Individuals and small teams |
| Team | $30 / seat / month | Everything in Pro, plus SSO, admin controls, and usage reporting. | Teams standardising on one assistant |
| API | Usage-based | All three models, prompt caching, batch inference, and spend caps. | Products built on the platform |
| Enterprise | Contact sales | Region pinning, zero-retention mode, audit log, named security contact. | Regulated and large organisations |
Prices shown are illustrative placeholder values for this design study.
The same model, in the place where the work already is.
Free tier, no card, and a usage policy that tells you plainly what the model will decline. If it fails your task, that is useful information and we would like to hear it.