Fronset Benchmark

Tencent Hy4 Preview at a glance

Good enough on 27/54 tasks at the 90% bar. Best value on 0 tasks. Best fit for: Financial Analysis & Trading Decisions, Infrastructure & Utility, Long-form Content Generation, Relevance, Classificatio

Provider
OpenRouter
Model name
tencent/hy4-preview
Cost mode:
Qualifies on
27 / 54 tasks (at 90% bar)
Best value on
0 tasks

Cost vs quality across all tasks

0%25%50%75%100%78910Quality score (7–10)Cost-efficiency vs best value (1.0 = best value)Voice Profile Generation — quality 9.47, 79.0x the cost of the best-value good-enough optionPublic Engagement Reply Review — quality 8.40, 4.8x the cost of the best-value good-enough optionModel Specific Prompt Adaptation — quality 8.91, 115.5x the cost of the best-value good-enough optionPromotional Message Generation — quality 8.23, 218.9x the cost of the best-value good-enough optionSection Prompt Generation — quality 8.74, 57.6x the cost of the best-value good-enough optionEvidence Grounded Claim Generation — quality 7.99, 35.4x the cost of the best-value good-enough optionAuthoritative Source Selection — quality 8.13, 2.2x the cost of the best-value good-enough optionBench Hypothesis Post Generation — quality 8.37, 223.9x the cost of the best-value good-enough optionContent To Image Prompt Generation — quality 8.56, 115.0x the cost of the best-value good-enough optionReport Executive Summary Generation — quality 7.87, 32.1x the cost of the best-value good-enough optionFactual Claim Refinement — quality 8.10, 91.5x the cost of the best-value good-enough optionStructured Content Summarization — quality 8.63, 102.2x the cost of the best-value good-enough optionThematic Topic Discovery — quality 7.87, 2.1x the cost of the best-value good-enough optionResearch Query Generation — quality 7.34, 101.1x the cost of the best-value good-enough optionLanguage Identification — quality 9.91, 76.3x the cost of the best-value good-enough optionBench Article Hypothesis Generation — quality 8.57, 9.8x the cost of the best-value good-enough optionPersona Policy Eligibility Check — quality 8.97, 81.8x the cost of the best-value good-enough optionProfile Pool Matching — quality 8.70, 47.8x the cost of the best-value good-enough optionTopic Section Assignment — quality 7.48, 67.5x the cost of the best-value good-enough optionSocial Post Portfolio Selection — quality 8.65, 205.6x the cost of the best-value good-enough optionProfiled Document Section Analysis — quality 8.98, 49.1x the cost of the best-value good-enough optionMulti Perspective Decision Synthesis — quality 8.30, 15.1x the cost of the best-value good-enough optionProfiled Document Structure Extraction — quality 8.64, 69.5x the cost of the best-value good-enough optionPublication Title Package Generation — quality 8.74, 106.0x the cost of the best-value good-enough optionVisual Theme Configuration Generation — quality 8.70, 75.0x the cost of the best-value good-enough optionProfiled Document Analysis — quality 8.84, 22.9x the cost of the best-value good-enough optionCatalyst And Scenario Analysis — quality 7.91, 45.1x the cost of 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
Thematic Topic DiscoveryTopic Organization & Clustering96%MEDIUM2.1x
Authoritative Source SelectionRelevance, Classification & Matching94%MEDIUM2.2x
Public Engagement Reply ReviewRelevance, Classification & Matching94%MEDIUM4.8x
Bench Article Hypothesis GenerationEvidence-Grounded Hypothesis Generation98%HIGH9.8x
Multi Perspective Decision SynthesisFinancial Analysis & Trading Decisions94%HIGH15x
Profiled Document AnalysisFinancial Analysis & Trading Decisions99%MEDIUM23x
Report Executive Summary GenerationContent Summarization & Synthesis91%MEDIUM32x
Evidence Grounded Claim GenerationContent Summarization & Synthesis94%MEDIUM35x
Catalyst And Scenario AnalysisFinancial Analysis & Trading Decisions90%HIGH45x
Profile Pool Matching bestRelevance, Classification & Matching100%MEDIUM48x
Profiled Document Section AnalysisFinancial Analysis & Trading Decisions91%HIGH49x
Section Prompt GenerationInfrastructure & Utility96%HIGH58x
Topic Section AssignmentTopic Organization & Clustering91%MEDIUM67x
Profiled Document Structure ExtractionStructured Data & Fact Extraction97%MEDIUM69x
Visual Theme Configuration GenerationLong-form Content Generation97%MEDIUM75x
Language IdentificationRelevance, Classification & Matching98%HIGH76x
Voice Profile GenerationLong-form Content Generation98%RANKED79x
Persona Policy Eligibility CheckRelevance, Classification & Matching97%MEDIUM82x
Factual Claim RefinementInfrastructure & Utility94%MEDIUM92x
Research Query GenerationInfrastructure & Utility93%HIGH101x
Structured Content SummarizationContent Summarization & Synthesis96%HIGH102x
Publication Title Package GenerationContent Summarization & Synthesis96%RANKED106x
Content To Image Prompt GenerationInfrastructure & Utility96%HIGH115x
Model Specific Prompt Adaptation bestInfrastructure & Utility100%MEDIUM116x
Social Post Portfolio SelectionRelevance, Classification & Matching99%RANKED206x
Promotional Message GenerationSocial & Promotional Content96%HIGH219x
Bench Hypothesis Post GenerationSocial & Promotional Content95%HIGH224x

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.