NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-25-01KYCRAQYZ84WNPZ5BJK61EMHN
Run Details
Model:
Claude Opus 5 (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
23.4s
Cost:
$0.030265
Provider:
Amazon Bedrock
Evaluation summary
Score
Evaluator
Details Link
81.27%
Matches word count
(details)
4.22%
Dialogue to Total Word Ratio
(details)
42.7446%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 30% words of dialogue between two thieves named Molly and Bill who are planning to break into a children's party to steal all the presents. Only include the prose of the scene without any markdown or commentary.
Model response
Messages
Meta
AI
The van smelled of old fries and older regret. Molly had spread the blueprints across the dashboard — not blueprints, technically, but a napkin sketch of the Pemberton back garden, complete with a bouncy castle rendered as a lopsided cloud and an arrow labeled LOOT ZONE. Bill studied it the way a man studies a tax form. "So the presents," he said, "are on a trestle table. Beside the piñata." "Beside and slightly behind. There's a tablecloth with cartoon dinosaurs on it, which works in our favor because dinosaurs are visually loud. Nobody looks at what's under a loud tablecloth." "That is not a real principle of anything." "It's a real principle of me." Molly had been casing the house since Tuesday, which mostly meant standing at the bus stop opposite it eating a scotch egg and taking notes. The birthday child was turning seven. His name was Rufus. Rufus had thirty-one invited guests, a hired magician, and — according to the laminated schedule taped inside the front window, because these people laminated things — a "PRESENT OPENING CEREMONY" at 3:45 sharp. Which meant the window was 2:00 to 3:44. Ninety-four minutes of unguarded plunder, guarded only by sugar-drunk children and a magician named The Astonishing Gary. "I want it on the record," Bill said, "that stealing from a seven-year-old is the lowest thing we've done." "Lower than the nunnery?" "The nuns had insurance." "Rufus has a barrister for a father and a mother who owns four wine fridges. Rufus has *contents cover*, Bill. Rufus will be re-gifted everything by Thursday and won't remember any of it by August. Meanwhile I am eating soup that comes in a sachet." Bill couldn't argue with the soup. He'd seen the soup. The plan, such as it was, had three moving parts. Molly would enter through the side gate at 2:10 disguised as a face painter, which required only a smock and the confidence of a woman carrying glitter. Bill would arrive at 2:25 as an entertainer's assistant, a role he had prepared for by buying a bow tie and practicing an expression of tolerant sorrow. At 2:40, Molly would announce a "special surprise for Rufus" in the front room, drawing the children away like a tide going out. That left the trestle table alone with Bill and four collapsible laundry bags. "And if the mother stays behind?" Bill asked. "She won't. She'll want photos." "And if The Astonishing Gary objects to an assistant he never hired?" "Then you say the word *union*, and you say it slowly, and you watch what happens to a man's face." Bill folded the napkin into his pocket with the care of an archivist. Outside, a bunch of helium balloons was being tied to a lamppost by someone in a linen shirt, and the sound of a distant bouncy castle inflating rose over the hedges like a great pneumatic sigh. "Ninety-four minutes," he said. "Ninety-four minutes." Molly checked her watch, then her glitter. "Try to look like you enjoy children." Bill attempted it. "Never mind," she said. "Look grieving. It's more convincing."
created_at
"2026-07-25T13:44:42.728Z"
7/25/2026, 1:44:42 PM
model
"anthropic/claude-opus-5-20260723"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3571
moderation_latency
(null)
generation_time
23319
tokens_prompt
75
tokens_completion
820
native_tokens_prompt
93
native_tokens_completion
1192
native_tokens_completion_images
(null)
native_tokens_reasoning
50
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"end_turn"
service_tier
"default"
usage
0.030265
router
(null)
provider_responses
0
endpoint_id
"76cb4608-f48c-483d-8da8-9957fb44244e"
id
"msg_011CdNsmUZW5pL3d9Vyssbr1"
is_byok
false
latency
1538
model_permaslug
"anthropic/claude-opus-5-20260723"
provider_name
"Amazon Bedrock"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.18.0; linux; x64))"
http_referer
(null)
request_id
"req-1784987082-Z17oVPhHNuOu2Vpxq4c9"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1784987082-m8wk1U0q4BXGWcHJRiP0"
upstream_id
"msg_011CdNsmUZW5pL3d9Vyssbr1"
total_cost
0.030265
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Amazon Bedrock"
response_cache_source_id
(null)
data_region
"global"
Evaluation details
Result
Evaluator
Details
Meta Data
81.27%
Matches word count
n/a
neededClean
false
words
512
4.22%
Dialogue to Total Word Ratio
Ratio: 37.96%, Deviation: 7.96%
neededClean
false
wordsTotal
519
wordsDialogue
197
42.7446%