NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-05-27-01KSMF7AC512DBJEZAN1A19ZNG
Run Details
Model:
Skyfall 36B V2
Test:
Text Replacement
Scenario:
Avoid said/asked/replied/answered
Duration:
7.9s
Cost:
$0.000682
Provider:
Parasail
Evaluation summary
Score
Evaluator
Details Link
8.33%
Key facts and content preserved
(details)
100.00%
Forbidden words eliminated
(details)
66.67%
Structural similarity to original
(details)
58.3333%
Overall
Model input
System
You are an expert prose editor. Whenever you're given text, rephrase it using the following instructions: <instructions>Remove all instances of 'said', 'asked', 'replied', and 'answered', replacing them with alternative dialogue attributions.</instructions> Imitate and keep the current writing style, and leave mannerisms, word choice and sentence structure intact. Keep the same stylistic writing choices. Use English spelling and grammar. Only return the rephrased text, nothing else.
User/Human
Text to rewrite: <text words="381"> The market square buzzed with the sound of a hundred voices. Elena pushed through the crowd, her satchel clutched against her chest. She knew the merchant would not wait — he never waited for anyone, least of all a girl from the outer ring. "You're late," Gregor said, without looking up from his ledger. He ran a thick finger down a column of numbers and frowned. Elena set the package on the counter. "The bridge was out. I had to take the long way around." "Not my problem." He unwrapped the cloth and held the crystal up to the light. It caught the sun and threw tiny rainbows across the stall's canvas roof. Gregor's frown deepened. "This isn't what I ordered." She had expected this. The old man always found something to complain about, whether the goods were perfect or not. Elena crossed her arms and stood her ground. "It's exactly what you ordered," she said. "Blue quartz, uncut, from the northern mines. I watched them pull it from the rock myself." Gregor turned the crystal slowly, examining every facet. He brought it close to his eye, then held it at arm's length. Elena thought he looked like a jeweler appraising a diamond, though Gregor dealt in stranger things than jewelry. He set the crystal down and sighed. "Fine. But next time, wrap it properly. The edges are exposed." He reached beneath the counter and produced a small leather pouch. The coins inside clinked as he slid it toward her. Elena scooped up the pouch and weighed it in her palm. It felt light. She loosened the drawstring and counted: eight silver crowns. They had agreed on ten. "We agreed on ten," she said flatly. "That was before the bridge fee went up." Gregor smiled — a thin, practiced smile that never reached his eyes. "Take it or leave it." She took it. She always took it. The outer ring didn't offer many choices, and Gregor knew that as well as she did. Elena walked out of the market and into the narrow lanes beyond. The sun hung low over the rooftops, and the shadows stretched long across the cobblestones. She slipped the pouch into her satchel and quickened her pace. If she hurried, she would reach the bridge before dark. </text>
Model response
Messages
Meta
AI
"What a headache," Gregor commented, without lifting his gaze from the ledger. His thick finger traversed down a row of coded numbers, and his scowl deepened. Gregory grasped the package and lifted the jewellry up to the sun. It sparkled, projecting rainbows across the canvas roofing above them. Gregor narrowed his eyes. "This isn't what I ordered." Elena had foreseen this. The dealer always located a flaw to gripe about, regardless of the merchandise's superior quality. Elena crossed her arms and held her ground. "I handed you exactly what you brokered for, and on time," she responded. "Clear, distinctive northern blue quartz. I observed it being dug up firsthand." Gregor rotated the stone carefully, analyzing it from all angles. He brought it up to his eyeline, then checked it from afar. Elena thought he looked like a jeweler assessing a diamond—even though Gregor peddled items infinitely more peculiar than jewelry. He set the gem down and exhaled. "Okay, here, but next time make sure it's wrapped securely, since the surfaces are vulnerable to damage." Gregor rummaged under the counter and retrieved a small leather sac. The jingle of coins filled the booth as he nudged it over to her. Elena snatched the sac and weighed it in her hand. It felt lighter than expected. She untied the drawstring and tallied the coins: eight silver crowns—two less than their originally negotiated sum. "But we agreed on ten," Elena declared flatly. "Well, the tolls just increased," Gregory remarked with a smile that never quite reached his eyes. "Take it or leave it." She took it. She always took it. The outer borough didn't provide many alternatives, and Gregor knew this as well as she did. Elena exited the market and entered the tight alleyways beyond. The sun had already started setting, and the shadows lengthened across the crumbling cobbled streets. Elena slipped the sac into her satchel and hastened her pace. Perhaps if she hurried, she could make it to the bridge before night.
created_at
"2026-05-27T10:20:59.665Z"
5/27/2026, 10:20:59 AM
model
"thedrummer/skyfall-36b-v2"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
169
moderation_latency
(null)
generation_time
7776
tokens_prompt
699
tokens_completion
519
native_tokens_prompt
605
native_tokens_completion
442
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
16
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
"stop"
service_tier
(null)
usage
0.00068155
router
(null)
provider_responses
0
endpoint_id
"1eb01ded-ae11-49e6-8aa6-3067584070bd"
id
"chatcmpl-ae9a206163a43903"
is_byok
false
latency
169
model_permaslug
"thedrummer/skyfall-36b-v2"
provider_name
"Parasail"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.15.0; linux; x64))"
http_referer
(null)
request_id
"req-1779877259-FX3iPOHcCPlohjm104MC"
session_id
(null)
api_type
"completions"
id
"gen-1779877259-FGdssNvNgiZroH9FvCTb"
upstream_id
"chatcmpl-ae9a206163a43903"
total_cost
0.00068155
cache_discount
0.0000048
upstream_inference_cost
0
provider_name
"Parasail"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
8.33%
Key facts and content preserved
1/12 anchor facts preserved
matched
1
total
12
details
0
id
"fact-01"
paragraphFound
true
preserved
false
1
id
"fact-02"
paragraphFound
true
preserved
true
2
id
"fact-03"
paragraphFound
true
preserved
false
3
id
"fact-04"
paragraphFound
true
preserved
false
4
id
"fact-05"
paragraphFound
true
preserved
false
5
id
"fact-06"
paragraphFound
true
preserved
false
6
id
"fact-07"
paragraphFound
true
preserved
false
7
id
"fact-08"
paragraphFound
true
preserved
false
8
id
"fact-09"
paragraphFound
true
preserved
false
9
id
"fact-10"
paragraphFound
true
preserved
false
10
id
"fact-11"
paragraphFound
true
preserved
false
11
id
"fact-12"
paragraphFound
false
preserved
false
100.00%
Forbidden words eliminated
All 3 forbidden word(s) eliminated
neededClean
false
totalOriginal
3
totalRemaining
0
perWord
0
word
"said"
original
3
remaining
0
1
word
"asked"
original
0
remaining
0
2
word
"replied"
original
0
remaining
0
3
word
"answered"
original
0
remaining
0
66.67%
Structural similarity to original
Words: 100%, Paragraphs: 0%, Sentences: 100%
wordCountScore
1
paragraphCountScore
0
sentenceCountScore
1
originalWords
381
outputWords
332
originalParagraphs
13
outputParagraphs
11
originalSentences
43
outputSentences
32
58.3333%