NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-07-09-01KX40YDTTVKRHFECS901Q598P
Run Details
Model:
GPT-5.6 Terra (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
14.1s
Cost:
$0.015855
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
56.40%
Dialogue to Total Word Ratio
(details)
28.2000%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 70% 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
“Children’s parties are ideal,” Molly whispered, peering through the hedge. “No guards, no alarms, and every victim is legally required to own a cake.” Bill adjusted his black turtleneck. “You said we were stealing presents, not cake.” “We steal the presents, Bill. The cake is operational morale.” “There are fourteen children in there.” “Thirteen. One is a cardboard pirate.” “It looks very convincing.” “That’s because it’s wearing an eye patch.” “So are we.” “We’re professionals. Eye patches create mystique.” “They create depth-perception problems.” Molly looked over the garden fence. Balloons bobbed above a table loaded with glittery packages. “There. The treasure pile.” Bill squinted. “Most of those are shaped like socks.” “Luxury socks. Probably with dinosaurs.” “And one appears to be a tambourine.” “Rare musical artifact.” “A plastic tambourine.” “Bill, we don’t judge loot by its composition. We judge it by how difficult it is to sell anonymously at a car boot sale.” “Who buys a used tambourine?” “Someone who hates silence.” A shriek erupted from the garden, followed by the unmistakable thump of a child falling into a bouncy castle. Bill flinched. “I don’t like this.” “You didn’t like the museum job.” “The museum had adults. Adults understand fear.” “Children understand fear.” “Not these ones. That tiny girl is holding a foam sword and shouting at a magician.” “She’s six.” “She’s feral.” Molly unfolded a crumpled sketch. “The plan is simple. You distract the entertainer. I collect the parcels. We exit through the side gate.” “The entertainer is a magician.” “Exactly. Ask him how he does the rabbit thing.” “And if he makes me disappear?” “Then I take your share.” “I don’t think we agreed on shares.” “We agreed on fifty-fifty.” “You said, ‘Don’t worry about it.’” “That is criminal shorthand for fifty-fifty.” Bill stared at the garden. “What if the birthday child notices?” “Which one is the birthday child?” “The one in the crown.” Molly paused. A small boy in a gold paper crown sat at the center of the chaos, solemnly watching the magician pull scarves from his sleeve. “He looks sad,” Bill said. “He looks suspicious.” “He’s holding a party blower like it’s a tiny trumpet of despair.” “Fine. We move quickly.” Bill reached into his jacket. “I brought disguises.” “Please tell me they aren’t eye patches.” He produced two cone-shaped party hats. Molly stared. “Those are worse.” “They make us look festive.” “They make us look like adults who have lost custody of their own judgment.” “Put one on.” “No.” “Molly.” She snatched a pink hat and jammed it onto her head. “If anyone asks, I’m a cousin.” “You can’t be a cousin. You’re wearing burglar gloves.” “They’re evening gloves.” “They have rubber grips.” “Very formal evening gloves.” They slipped through the gate. A child immediately ran up to Bill. “Are you here for the treasure hunt?” she demanded. Bill froze. “Possibly.” “Good. We need another pirate.” “I’m not dressed as a pirate.” “You have an eye patch.” “It’s professional mystique.” The girl shoved a plastic cutlass into his hand. “You guard the crocodile.” “What crocodile?” A green inflatable crocodile rose from behind a paddling pool. Molly sidled toward the presents. “Keep her occupied.” “How?” “Talk pirate.” Bill swallowed. “Arrr?” The children cheered. Molly grabbed a parcel, then read the tag. “To Oliver, from Gran.” She picked up another. “To Oliver, from Auntie Sue.” A third. “To Oliver, from Mum and Dad.” Bill was surrounded by children chanting, “Crocodile! Crocodile!” Molly looked at the sad boy in the crown. Then at the pile of gifts. She sighed. “Bill.” “What?” “Abort.” “Why?” “Because apparently we’re stealing joy from a child named Oliver.” Bill lowered the foam sword. “Is that bad?” “Yes.” “Even if there’s a tambourine?” “Especially the tambourine.” Bill smiled at the children. “Arrr. The crocodile has surrendered.” Oliver finally smiled. Molly slipped a slice of cake into her handbag. “Operational morale,” she muttered.
created_at
"2026-07-09T18:06:24.61Z"
7/9/2026, 6:06:24 PM
model
"openai/gpt-5.6-terra-20260709"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1752
moderation_latency
(null)
generation_time
13345
tokens_prompt
77
tokens_completion
1302
native_tokens_prompt
66
native_tokens_completion
1046
native_tokens_completion_images
0
native_tokens_reasoning
103
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
"completed"
service_tier
"default"
usage
0.015855
router
(null)
provider_responses
0
endpoint_id
"3cc89000-ae34-4dde-9c7d-5ce009c8af0b"
id
"resp_03e904f2bcf96971016a4fe320ae14819390b20b7226101997"
is_byok
false
latency
259
model_permaslug
"openai/gpt-5.6-terra-20260709"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.18.0; linux; x64))"
http_referer
(null)
request_id
"req-1783620384-uzpjog1TivgHWdRAdLxm"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1783620384-qY1KvSm8SED9z0BBmIoy"
upstream_id
"resp_03e904f2bcf96971016a4fe320ae14819390b20b7226101997"
total_cost
0.015855
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
response_cache_source_id
(null)
data_region
"global"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
649
56.40%
Dialogue to Total Word Ratio
Ratio: 66.62%, Deviation: 3.38%
neededClean
false
wordsTotal
653
wordsDialogue
435
28.2000%