NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37NVSGJYS9DQF8X7T23SKFD
Run Details
Model:
Claude Opus 5.5 (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
30.8s
Cost:
$0.050000
Provider:
Claude Platform on AWS
Evaluation summary
Score
Evaluator
Details Link
9.61%
Matches word count
(details)
69.15%
Dialogue to Total Word Ratio
(details)
39.3807%
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 back booth of Denny's smelled like syrup and poor decisions, which was exactly why Molly Pratchett had chosen it. She unrolled the blueprint across the table, pinning one corner with the ketchup and another with Bill's elbow, which he had not offered. Bill Hoskins squinted at the drawing. It was crayon. It showed a lopsided house, a sun wearing sunglasses, and a stick figure labeled ME. "Where did you get this?" "The birthday girl drew it for show-and-tell. Her teacher posted it online." Molly tapped the sun. "Never underestimate a second-grader's grasp of floor plans." Bill had been a thief for nineteen years. He had cracked safes in Geneva, lifted a Fabergé egg from a moving yacht, and once stolen a racehorse by convincing it he was its mother. He had never robbed a seven-year-old, and something in his stomach, possibly the Grand Slam, was objecting. "Molly, it's a kid's party." "It's a kid's party in Greenwich, Bill. Those gift bags alone are worth more than your car." She wasn't wrong. Bill's car was a 2004 Corolla with a raccoon living in the trunk rent-free. Still, he leaned back and folded his arms, the universal posture of a man about to lose an argument. "Walk me through it." Molly's eyes lit up the way they did whenever someone gave her permission to be terrible. She produced a stack of index cards, color-coded and laminated, because Molly laminated everything, including, once, a restraining order. "Two o'clock, the magician arrives. Except the magician is you." "I don't know any magic." "You don't need magic. You need a cape and the confidence of a man who has never been loved. While you distract forty children by pulling a quarter from somewhere, I slip into the gift room disguised as a party balloon." Bill waited for the part where she explained how a person disguised herself as a balloon. It did not come. He considered asking, then decided he'd rather not know. Molly simply moved to the next card, which showed a getaway route drawn around a bouncy castle. "What about the parents?" "Mimosas. Endless. They won't notice if the house itself leaves." "And the kids?" Molly hesitated, and for the first time since Bill had known her, a flicker of genuine fear crossed her face. She lowered her voice. "The kids are the problem. Children see everything. They have no inhibitions, no social contract, and they bite." Bill thought of his nephew, who had once identified him in a lineup of department store Santas purely by smell. He nodded gravely. "So we need a distraction for the distraction." "Now you're thinking like a professional." Outside, the rain began to fall, and somewhere in Greenwich a little girl named Penelope was sleeping peacefully, unaware that in thirty-six hours she would become the most dangerous adversary either of them had ever faced.
created_at
"2026-09-23T17:43:52.611Z"
9/23/2026, 5:43:52 PM
model
"anthropic/claude-opus-5.5-20260921"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
5028
moderation_latency
(null)
generation_time
30754
tokens_prompt
75
tokens_completion
1106
native_tokens_prompt
95
native_tokens_completion
2481
native_tokens_completion_images
(null)
native_tokens_reasoning
1498
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
0.05
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790185432-XRNO6SiIXv6KsOyAY3Fk"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790185432-0J3KX1B980BjIVL3sqK1"
upstream_id
"msg_011CfLnYCCz9y3R83jWAHGLx"
provider_responses
0
endpoint_id
"3a2388bc-3740-4e64-a1f5-4e301726b6b9"
id
"msg_011CfLnYCCz9y3R83jWAHGLx"
is_byok
false
latency
1566
model_permaslug
"anthropic/claude-opus-5.5-20260921"
provider_name
"Claude Platform on AWS"
status
200
total_cost
0.05
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
9.61%
Matches word count
n/a
neededClean
false
words
478
69.15%
Dialogue to Total Word Ratio
Ratio: 32.72%, Deviation: 2.72%
neededClean
false
wordsTotal
486
wordsDialogue
159
39.3807%