NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37PMGV2XTJ488XC4V6XTAR4
Run Details
Model:
Claude Opus 5.5 (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
42.7s
Cost:
$0.079500
Provider:
Claude Platform on AWS
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
98.86%
Dialogue to Total Word Ratio
(details)
99.4313%
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 cold coffee and ambition. Molly spread the blueprint across the dashboard, smoothing its creases with the reverence of a surgeon laying out instruments. It was not a blueprint. It was a birthday invitation shaped like a dinosaur, and someone had drawn the floor plan on the back in purple crayon. Bill squinted at it through the fog of his own breath. He had been a getaway driver for eleven years and had never once gotten away, which Molly considered a statistical achievement. "Walk me through it again," he said. "Saturday, two o'clock. The Pemberton twins turn seven. Forty guests, one bouncy castle, and a present table groaning under the weight of consumer goods." Molly tapped the crayon rectangle labeled PRESANTS. "That's our vault." "And security?" "Two parents, both exhausted. One grandmother with a camcorder. And a clown." Bill went pale. "You never said there'd be a clown. Clowns have eyes that follow you, Molly. Like portraits." Molly sighed. Every crew had a weakness. Hers happened to be six foot four, wore a balaclava to the grocery store, and had been emotionally scarred by a juggler at a county fair in 1994. She reached into the glovebox and produced two folded costumes, which unfurled into a pair of enormous foam bananas. "We go in as entertainment," she explained. "Nobody questions a banana. The banana is a trusted fruit." Bill held his costume at arm's length, as though it might bite. "What do bananas do at a party?" "Dance. Sing. Distract. While the children are dazzled, I load the gifts into the piƱata bag, and you create a diversion." "What kind of diversion?" "You'll know it when you see it. Improvise." This was a mistake, and a part of Molly knew it even then. The last time Bill had improvised, he had set off a fire alarm, apologized to the fire alarm, and then waited politely for the police to arrive so he could explain himself. He had a deep, almost spiritual respect for authority, which was unusual in a career criminal. Still, the numbers were irresistible. The Pembertons were the kind of wealthy that bought seven-year-olds drones with better cameras than most banks. Somewhere on that table sat a limited-edition robot dog worth more than the van, and Molly intended to own it by sundown. Bill turned the invitation over and read the front aloud, moving his lips slowly. Then he stopped. "Molly," he said. "It says here there's a goody bag for every guest." "So? Goody bags are cheap plastic whistles, Bill." "So if we're guests, we get a goody bag. Legally. Nobody chases you for a goody bag." She stared at him. He stared back, a man who had just discovered honest income. Outside, a streetlight flickered. Molly folded the banana suits with great dignity and placed them back in the glovebox. Some heists were about the score. Others were about the principle. And this one, she decided grimly, was about the robot dog.
created_at
"2026-09-23T17:57:22.937Z"
9/23/2026, 5:57:22 PM
model
"anthropic/claude-opus-5.5-20260921"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
4459
moderation_latency
(null)
generation_time
42616
tokens_prompt
75
tokens_completion
1434
native_tokens_prompt
95
native_tokens_completion
3956
native_tokens_completion_images
(null)
native_tokens_reasoning
2940
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
0.0795
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790186242-mTrf06cPtvJtJnV7nGuF"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790186242-a9rpROlTRAHPeQuvoZ24"
upstream_id
"msg_011CfLoZvhknyD56uX6eaH3p"
provider_responses
0
endpoint_id
"3a2388bc-3740-4e64-a1f5-4e301726b6b9"
id
"msg_011CfLoZvhknyD56uX6eaH3p"
is_byok
false
latency
1262
model_permaslug
"anthropic/claude-opus-5.5-20260921"
provider_name
"Claude Platform on AWS"
status
200
total_cost
0.0795
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Claude Platform on AWS"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
499
98.86%
Dialogue to Total Word Ratio
Ratio: 30.48%, Deviation: 0.48%
neededClean
false
wordsTotal
502
wordsDialogue
153
99.4313%