NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-01-01KWEABQFZ4KS2DRV45MKTRPQW
Run Details
Model:
Claude Sonnet 5 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
17.6s
Cost:
$0.011226
Provider:
Amazon Bedrock
Evaluation summary
Score
Evaluator
Details Link
75.16%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
37.5778%
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
Molly sat at the greasy diner table, spreading out a hand-drawn map that looked suspiciously like it had been sketched by a five-year-old. Which, in fairness, it had been—she'd swiped it from her nephew's backpack that morning. "This is the floor plan of the Hendricks house," she said, tapping a crayon rendering of what might have been a house or possibly a very angry potato. "Party starts at two. Bounce castle here, cake table there, and the presents—" she jabbed the map with authority, "—right by the sliding glass door." Bill leaned back in the booth, arms crossed. "You're planning a heist based on a drawing made by a child." "A very observant child." "It says 'Timmy is a poopy head' in the corner." "Every good map has annotations." Bill rubbed his temples. Twenty years in the game, and somehow he'd ended up planning a job with a woman who thought crayon cartography counted as reconnaissance. "Molly, we are robbing a seven-year-old's birthday party." "We're liberating unopened presents from children who don't even appreciate them yet. Half these kids will forget what they got by Tuesday." "That's not the moral high ground you think it is." "I never said it was moral. I said it was profitable." She pulled out a second sheet—an actual invoice, water-stained but legible. "Timmy's getting a drone, a bike, two video game consoles, and God knows what else. That's easily eight hundred dollars in loot, Bill. Eight hundred dollars, sitting in shiny wrapping paper, guarded by a bouncer dressed as a clown." "The bouncer is the clown?" "Multitasking. Budget cuts." Bill exhaled slowly, the exhale of a man who had made many bad decisions in his life and sensed another queuing up politely behind the others. "And how exactly do two adults waltz into a child's birthday party without every parent in a fifty-foot radius immediately calling the police?" Molly grinned, far too pleased with herself, and reached into her bag. She produced two costumes—one a slightly moth-eaten bear suit, the other something resembling a rejected mascot for a regional insurance company. "We're the entertainment." "Absolutely not." "You'll be BooBoo the Bear. I'll be Sparkles the Unicorn. We do a magic show, keep the kids distracted in the backyard, and while their parents are busy filming little Timmy sawed in half—not literally, Bill, calm down—we slip inside and grab the goods." "This plan has at least fourteen ways to go wrong." "Sure. But none of them involve armed guards or laser grids, which, frankly, is an upgrade from our usual Tuesdays." Bill stared at the bear costume, at its slightly unsettling grin, and then at Molly, who was already pulling the unicorn horn from her bag like it was a crown of victory. Somewhere deep down, past the exhaustion and the very real fear of felony charges involving a bounce castle, a small, treacherous part of him thought: *this might actually work.* "Fine," he muttered. "But I'm not doing balloon animals." "Deal. I'll handle the animals. You just look menacing in fur."
created_at
"2026-07-01T07:47:40.171Z"
7/1/2026, 7:47:40 AM
model
"anthropic/claude-sonnet-5-20260630"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
6415
moderation_latency
(null)
generation_time
17617
tokens_prompt
77
tokens_completion
817
native_tokens_prompt
93
native_tokens_completion
1104
native_tokens_completion_images
(null)
native_tokens_reasoning
18
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
0.011226
router
(null)
provider_responses
0
endpoint_id
"37b16f45-eeae-45f3-9b85-2137f7c6d61b"
id
"msg_01KVhC8vkgc5Lcc1vHYERVqh"
is_byok
false
latency
1573
model_permaslug
"anthropic/claude-sonnet-5-20260630"
provider_name
"Amazon Bedrock"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.17.0; linux; x64))"
http_referer
(null)
request_id
"req-1782892060-LbB1AtUVcxSTn1bWImUW"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1782892060-s2AKhHt1YTN4a8dkt8Lp"
upstream_id
"msg_01KVhC8vkgc5Lcc1vHYERVqh"
total_cost
0.011226
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
75.16%
Matches word count
n/a
neededClean
false
words
513
0.00%
Dialogue to Total Word Ratio
Ratio: 53.17%, Deviation: 23.17%
neededClean
false
wordsTotal
521
wordsDialogue
277
37.5778%