NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37PW99AAS916105FTR3H5DG
Run Details
Model:
Claude Opus 5.5 (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
39.2s
Cost:
$0.066420
Provider:
Claude Platform on AWS
Evaluation summary
Score
Evaluator
Details Link
93.65%
Matches word count
(details)
92.34%
Dialogue to Total Word Ratio
(details)
92.9968%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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
Molly spread the blueprint across the hood of the van, pinning one corner with a thermos and the other with Bill's elbow, which he had not volunteered. They were parked behind a laundromat on Elm Street, the traditional headquarters of every great criminal enterprise that couldn't afford rent. "This is the Hendersons' backyard," she said. "Saturday, two o'clock. Timmy turns seven." Bill squinted. "That's a drawing of a bouncy castle." "It's a schematic of a bouncy castle." "You did it in crayon." "I did it in character. Focus." A breeze lifted the paper, and Bill slapped it back down, smearing a purple clown into what now resembled a bruise with feet. Molly let it go. Professionals adapted. "The presents go on a folding table by the patio," she continued. "Forty kids, forty gifts, minimum. At three-fifteen, the dad brings out the cake. Every eye in that yard goes to the candles." "And that's when we grab the loot." "That's when you grab the loot. I'll be the distraction." Bill folded his arms. "What kind of distraction?" Molly reached into the van and pulled out a garment bag with the grim ceremony of a surgeon unveiling a scalpel. She unzipped it. Inside hung a rainbow wig, a polka-dot jumpsuit, and a pair of shoes long enough to be classified as watercraft. "Absolutely not," said Bill. "It's for me. Mister Giggles." "You're going to be a clown called Mister Giggles." "I'm going to be a clown called Mister Giggles who shows up uninvited and does forty-five minutes of balloon animals. Nobody questions a clown, Bill. It's the perfect crime." Bill pictured Molly, five-foot-one and famously allergic to joy, twisting a poodle out of latex. The image refused to load. "What if they ask you to make a giraffe?" "Everything's a giraffe if you're confident." She demonstrated with an invisible balloon, producing an invisible something that she clearly believed in. Bill looked unconvinced, which was his natural resting face, so Molly pressed on. "You'll be in the hedge by the fence. When the singing starts, you crawl out, load the bags, and roll them to the van." "There's a problem." "There's no problem." "There's a big problem." Bill pointed at the corner of the blueprint, where a small, furious stick figure had been drawn beside the gate. "What's that?" Molly's confidence flickered. "That's Grandma Henderson." "Why is she holding a sword?" "It's a cane. I couldn't draw a cane." "Why is the cane on fire?" "Because she's seventy-eight, she does aqua aerobics, and last year she tackled a mailman for leaving a package in the rain." Molly exhaled. "She's the most dangerous person in the neighborhood. Possibly the state." Bill stared at the flaming stick figure for a long moment. Somewhere down the street, an ice cream truck played a tinny melody that sounded, to his ears, like a funeral march. "I want hazard pay," he said. "You'll get first pick of the loot." "The loot is Lego sets and Beyblades." "Then pick well," said Molly, and zipped the clown suit shut.
created_at
"2026-09-23T18:01:37.336Z"
9/23/2026, 6:01:37 PM
model
"anthropic/claude-opus-5.5-20260921"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
5274
moderation_latency
(null)
generation_time
39150
tokens_prompt
75
tokens_completion
1217
native_tokens_prompt
95
native_tokens_completion
3302
native_tokens_completion_images
(null)
native_tokens_reasoning
2177
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
0.06642
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790186497-65UJUi9rQsPWHBds2nDt"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790186497-233DMjlzjxFntKXFHkNa"
upstream_id
"msg_011CfLotgQJnpQPbbmhtQmxV"
provider_responses
0
endpoint_id
"3a2388bc-3740-4e64-a1f5-4e301726b6b9"
id
"msg_011CfLotgQJnpQPbbmhtQmxV"
is_byok
false
latency
1322
model_permaslug
"anthropic/claude-opus-5.5-20260921"
provider_name
"Claude Platform on AWS"
status
200
total_cost
0.06642
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
93.65%
Matches word count
n/a
neededClean
false
words
509
92.34%
Dialogue to Total Word Ratio
Ratio: 51.26%, Deviation: 1.26%
neededClean
false
wordsTotal
515
wordsDialogue
264
92.9968%