NC Bench
Overview
About
Tests
Models
Model
Claude Haiku 4.5
Test
Dialogue tags
Scenario
Write 500 words with 30% dialogue
Run ID
2026-02-08-01KGYYY2HAZYHC3RQ9J49FNWDB
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 Birthday Job Molly spread the floor plan across the sticky card table, careful not to let her sleeve touch the mysterious residue that covered every surface of Bill's apartment. She squinted at the crude drawing of Timmy Henderson's house, which she'd obtained by pretending to be a wedding planner who'd gotten the wrong address. "So we go in through the kitchen window," she said, tapping the diagram with her pen. "You're sure about the timing?" Bill nodded, adjusting his binoculars even though they were indoors. He'd never fully explained why he wore them constantly, and Molly had stopped asking around the time he tried to use them to read her grocery list from across the room. "Party starts at two. Peak chaos is two-thirty, when they do the piñata. Everyone's focused on a blindfolded seven-year-old swinging a stick around. Security will be at minimum." "Security?" Molly looked up. "Bill, it's a children's birthday party, not a bank." "You can never be too careful," he said gravely. "Last week I was casing a juice box shipment, and there was an actual German Shepherd guarding the loading dock. A dog, Molly. Completely unexpected." She decided not to point out that the dog was probably attached to a house nearby and not actually guarding juice boxes. Bill operated in his own logical universe, and it was oddly effective. "What about the presents?" Bill asked, pulling out a notebook filled with increasingly illegible sketches. "I count forty-three wrapped boxes. Maybe forty-four. I got distracted counting and started again three times." "We don't need all of them," Molly said practically. "Just the big ones. The really expensive stuff—those are usually by the fireplace." "Right, right. The north wall of the living room. I've memorized the layout." Bill tapped his temple proudly. "Photographic memory. Well, not exactly photographic. More like a memory that occasionally remembers things, but not always the same things, if you know what I mean." "I really don't." "The video game system is priority one," Bill continued, apparently unaffected by her confusion. "And that bicycle. Who gets a bicycle at an indoor party? That's arrogance. That's a statement." "It's a birthday party, Bill. Kids get presents. That's the whole thing." "Exactly my point." Molly studied him carefully, wondering again why she kept working with this man. He'd once convinced a museum guard that he was a plumbing inspector by drawing pipes on his hands with marker. The job had taken three months to plan and forty-five minutes to execute. "How are we getting out?" she asked. Bill's face fell. He flipped through his notebook desperately, page after page of sketches and notes. Finally, he looked up. "I didn't think that far ahead." Molly sighed. She never did either, but at least she had the decency to pretend she had a plan.
created_at
"2026-02-08T15:46:50.688286+00:00"
Sunday, February 8th, 2026 at 3:46:50 PM GMT+00:00
model
"anthropic/claude-4.5-haiku-20251001"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
931
moderation_latency
185
generation_time
7941
tokens_prompt
61
tokens_completion
607
native_tokens_prompt
75
native_tokens_completion
690
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.003525
router
(null)
provider_responses
0
id
"20d2fa69-7827-4ba9-8836-1e851ef6f1f4"
status
200
is_byok
false
latency
931
endpoint_id
"d000261c-5aeb-47d1-ac87-ad64dbe1cdb6"
provider_name
"Amazon Bedrock"
model_permaslug
"anthropic/claude-4.5-haiku-20251001"
api_type
"completions"
id
"gen-1770565601-hN2p3wtiwQiec1fGr7Gl"
upstream_id
"20d2fa69-7827-4ba9-8836-1e851ef6f1f4"
total_cost
0.003525
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Amazon Bedrock"
Evaluation details
Result
Evaluator
Details
Meta Data
6.0907%
Matches word count
n/a
neededClean
false
words
477
0.0004%
Dialogue to Total Word Ratio
Ratio: 45.70%, Deviation: 15.70%
neededClean
false
wordsTotal
477
wordsDialogue
218
3.0456%