Fronset Benchmark

GPT-6 Luna at a glance

Good enough on 24/54 tasks at the 90% bar. Best value on 15 tasks. Best fit for: Infrastructure & Utility, Content Summarization & Synthesis.

Provider
OpenAI
Model name
gpt-6-luna
Cost mode:
Qualifies on
24 / 54 tasks (at 90% bar)
Best value on
15 tasks

Cost vs quality across all tasks

0%25%50%75%100%678910Quality score (6–10)Cost-efficiency vs best value (1.0 = best value)Claim Continuity Classification — quality 8.04, this model IS the best-value good-enough option★Voice Profile Generation — quality 8.86, this model IS the best-value good-enough option★Model Specific Prompt Adaptation — quality 8.43, this model IS the best-value good-enough option★Section Prompt Generation — quality 8.49, this model IS the best-value good-enough option★Evidence Grounded Claim Generation — quality 7.97, this model IS the best-value good-enough option★Atomic Fact Claim Extraction — quality 8.10, this model IS the best-value good-enough option★Generated Content Relevance Scoring — quality 8.42, 2.4x the cost of the best-value good-enough optionMetadata Paragraph Rewriting — quality 7.81, 1.4x the cost of the best-value good-enough optionStructured Output Extraction — quality 9.11, 1.5x the cost of the best-value good-enough optionBatch Text Translation — quality 8.63, 1.6x the cost of the best-value good-enough optionBench Hypothesis Post Generation — quality 8.53, this model IS the best-value good-enough option★Content To Image Prompt Generation — quality 8.75, 1.1x the cost of the best-value good-enough optionReport Executive Summary Generation — quality 8.06, this model IS the best-value good-enough option★Factual Claim Refinement — quality 8.62, this model IS the best-value good-enough option★Structured Content Summarization — quality 8.54, this model IS the best-value good-enough option★Research Query Generation — quality 7.09, this model IS the best-value good-enough option★Language Identification — quality 10.00, 1.4x the cost of the best-value good-enough optionContent Set Relevance Scoring — quality 6.30, 1.5x the cost of the best-value good-enough optionSocial Post Portfolio Selection — quality 8.55, this model IS the best-value good-enough option★Profiled Document Section Analysis — quality 9.54, this model IS the best-value good-enough option★Topic Cluster Labeling — quality 8.32, 1.0x the cost of the best-value good-enough optionPublication Title Package Generation — quality 8.61, 1.4x the cost of the best-value good-enough optionVisual Theme Configuration Generation — quality 8.59, this model IS the best-value good-enough option★Catalyst And Scenario Analysis — quality 8.60, this model IS the best-value good-enough option★

within ~1.3× of the best-value model · 1.3–2× · >2× · ★ this model is the best-value pick on that task. Top-right = best quadrant. Only tasks where this model qualifies at the 90% bar are plotted.

Per-task breakdown

TaskCategoryQuality (% of best)ConfidenceOverpay
Structured Content Summarization ★Content Summarization & Synthesis95%HIGHbest value
Section Prompt Generation ★Infrastructure & Utility93%MEDIUMbest value
Evidence Grounded Claim Generation ★Content Summarization & Synthesis93%MEDIUMbest value
Visual Theme Configuration Generation ★Long-form Content Generation96%RANKEDbest value
Model Specific Prompt Adaptation ★Infrastructure & Utility95%MEDIUMbest value
Atomic Fact Claim Extraction ★ bestStructured Data & Fact Extraction100%MEDIUMbest value
Factual Claim Refinement ★Infrastructure & Utility100%HIGHbest value
Voice Profile Generation ★Long-form Content Generation92%RANKEDbest value
Claim Continuity Classification ★Relevance, Classification & Matching92%RANKEDbest value
Profiled Document Section Analysis ★Financial Analysis & Trading Decisions97%HIGHbest value
Report Executive Summary Generation ★Content Summarization & Synthesis94%MEDIUMbest value
Research Query Generation ★Infrastructure & Utility90%HIGHbest value
Catalyst And Scenario Analysis ★Financial Analysis & Trading Decisions98%MEDIUMbest value
Social Post Portfolio Selection ★Relevance, Classification & Matching98%HIGHbest value
Bench Hypothesis Post Generation ★Social & Promotional Content97%RANKEDbest value
Topic Cluster LabelingTopic Organization & Clustering93%HIGH1x
Content To Image Prompt GenerationInfrastructure & Utility98%RANKED1.1x
Publication Title Package GenerationContent Summarization & Synthesis94%MEDIUM1.4x
Metadata Paragraph RewritingInfrastructure & Utility93%RANKED1.4x
Language IdentificationRelevance, Classification & Matching99%RANKED1.4x
Content Set Relevance ScoringRelevance, Classification & Matching90%MEDIUM1.5x
Structured Output ExtractionStructured Data & Fact Extraction92%HIGH1.5x
Batch Text TranslationInfrastructure & Utility93%HIGH1.6x
Generated Content Relevance ScoringRelevance, Classification & Matching94%MEDIUM2.4x

Overpay — how much more you pay by running this model instead of the best-value model that clears the quality bar on that task (marked ★). "16x" means you overpay 16× — the same output for 16× the best-value good-enough option; ★ means this model is that option (no overpayment). Confidence — how sure we are about the quality score (more judgments + more agreement = higher confidence): RANKED many independent judges scored this model's outputs and their agreement is very high (most confident) — HIGH many judges have scored it and they mostly agree (well-pinned) — MEDIUM enough judges have weighed in to publish, but they disagree more than we'd like (treat with a small grain of salt). LOW-confidence cells are hidden everywhere on the site. See the methodology for the exact thresholds.