Root cause: sequential execution of "messages.create"

3 "anthropic.messages.create" spans ran one after another — each started only after the previous one finished. Total: 5.2s, slowest single call: 2.0s. Running them concurrently saves 3.3s — a 2.7× speedup.

Now (sequential)After fix (parallel)

5.2s

→

~2.0s

Fix: run calls in parallel

fix.py

from concurrent.futures import ThreadPoolExecutor

# Before — sequential, each call blocks the next:
for item in items:
    result = process(item)

# After — all calls run concurrently:
with ThreadPoolExecutor(max_workers=8) as pool:
    results = list(pool.map(process, items))

# Cost unchanged. Output unchanged. Time = slowest single call.

Waterfall

workflow › execute

workflow.execute

user: u_heavy_28

5.89s

anthropic › messages › create

anthropic.messages.create

claude-sonnet-4-63.0k tok$0.024faithful 0.95user: u_heavy_28

1.96s

Input

Summarize the Q3 earnings report attached. Focus on revenue, margin, and guidance.

Output

(answer to: Summarize the Q3 earnings report attache…)

anthropic › messages › create

anthropic.messages.create

claude-sonnet-4-62.7k tok$0.021faithful 0.97user: u_heavy_28

1.56s

Input

Given the customer ticket below, draft a refund response that follows policy P-204.

Output

(answer to: Given the customer ticket below, draft a…)

anthropic › messages › create

anthropic.messages.create

claude-sonnet-4-62.5k tok$0.019faithful 0.96user: u_heavy_28

1.72s

Input

Translate the user manual section to Japanese, preserving the table structure.

Output

(answer to: Translate the user manual section to Jap…)

Tool: calculator

tool.calculator

user: u_heavy_28

560ms

Faithfulness · anthropic.messages.create0.25

1 supported · 1 contradicted · 2 unsupported of 4 claims

supported
Day 1 plan fits under the $200/day soft cap.
contradicted
Belém Tower opens at 9am.
unsupported
Time Out Market lunch costs $18.
unsupported
Castelo dinner costs $42.