The Right-Sized LLM Benchmark Cheapest LLM that's good enough.

Tencent Hy3 at a glance

Good enough on 16/52 tasks at the 90% bar. Best value on 4 tasks. Best fit for: Content Summarization & Synthesis, Topic Organization & Clustering. Doesn't qualify on any: Financial Analysis & Trading

Provider
OpenRouter
Model name
tencent/hy3
Cost mode:
Qualifies on
16 / 52 tasks (at 90% bar)
Best value on
4 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)Topic Sequence Ordering — quality 8.19, this model IS the best-value good-enough option★Content Summarization — quality 8.73, 1.9x the cost of the best-value good-enough optionActivity Feed Blurb Generation — quality 8.81, 4.4x the cost of the best-value good-enough optionSubstack Newsletter — quality 8.52, 4.0x the cost of the best-value good-enough optionLanguage Detection — quality 9.90, 3.8x the cost of the best-value good-enough optionPublication Title Generation — quality 8.34, 3.5x the cost of the best-value good-enough optionLanding Page Section Generation — quality 9.06, this model IS the best-value good-enough option★Topic Cluster Naming — quality 8.05, 1.1x the cost of the best-value good-enough optionEngagement Triage — quality 8.39, 1.8x the cost of the best-value good-enough optionPrompt Adaptation — quality 8.75, this model IS the best-value good-enough option★Onboarding Chapter Outline Generation — quality 8.91, 1.6x the cost of the best-value good-enough optionX Post Selection — quality 8.11, 6.0x the cost of the best-value good-enough optionClaim Refinement — quality 8.47, 3.2x the cost of the best-value good-enough optionImage Prompt Generation — quality 8.33, 2.9x the cost of the best-value good-enough optionEngagement Reply Draft — quality 7.76, this model IS the best-value good-enough option★Topic-to-Section Assignment — quality 8.19, 2.8x 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
Topic Sequence Ordering ★Topic Organization & Clustering91%MEDIUMbest value
Prompt Adaptation ★Infrastructure & Utility98%RANKEDbest value
Engagement Reply Draft ★Social & Promotional Content92%MEDIUMbest value
Landing Page Section Generation ★Long-form Content Generation92%HIGHbest value
Topic Cluster NamingTopic Organization & Clustering92%RANKED1.1x
Onboarding Chapter Outline GenerationLong-form Content Generation91%MEDIUM1.6x
Engagement TriageRelevance, Classification & Matching94%MEDIUM1.8x
Content SummarizationContent Summarization & Synthesis92%HIGH1.9x
Topic-to-Section AssignmentTopic Organization & Clustering91%MEDIUM2.8x
Image Prompt GenerationInfrastructure & Utility91%HIGH2.9x
Claim Refinement bestInfrastructure & Utility100%HIGH3.2x
Publication Title GenerationContent Summarization & Synthesis93%RANKED3.5x
Language DetectionRelevance, Classification & Matching98%RANKED3.8x
Substack NewsletterLong-form Content Generation92%HIGH4x
Activity Feed Blurb Generation bestSocial & Promotional Content100%RANKED4.4x
X Post SelectionRelevance, Classification & Matching93%HIGH6x

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.