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

Claude Haiku 4.5 at a glance

Good enough on 9/52 tasks at the 90% bar. Best value on 0 tasks. Doesn't qualify on any: Structured Data & Fact Extraction, Social & Promotional Content, Infrastructure & Utility.

Provider
Anthropic
Model name
claude-haiku-4-5
Cost mode:
Qualifies on
9 / 52 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)Substack Newsletter — quality 9.04, 11.6x the cost of the best-value good-enough optionTrading Recommendation — quality 8.01, 7.9x the cost of the best-value good-enough optionLanguage Detection — quality 9.94, 15.7x the cost of the best-value good-enough optionExecutive Summary Generation — quality 7.67, 8.6x the cost of the best-value good-enough optionPublication Title Generation — quality 8.16, 7.6x the cost of the best-value good-enough optionTopic Cluster Naming — quality 8.41, 3.1x the cost of the best-value good-enough optionEngagement Triage — quality 8.71, 11.2x the cost of the best-value good-enough optionX Post Selection — quality 8.04, 7.6x the cost of the best-value good-enough optionAuthor Voice Generation — quality 8.93, 19.0x 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 Cluster NamingTopic Organization & Clustering96%HIGH3.1x
X Post SelectionRelevance, Classification & Matching92%MEDIUM7.6x
Publication Title GenerationContent Summarization & Synthesis91%RANKED7.6x
Trading RecommendationFinancial Analysis & Trading Decisions92%MEDIUM7.9x
Executive Summary GenerationContent Summarization & Synthesis97%HIGH8.6x
Engagement TriageRelevance, Classification & Matching98%RANKED11x
Substack NewsletterLong-form Content Generation98%RANKED12x
Language DetectionRelevance, Classification & Matching99%RANKED16x
Author Voice GenerationLong-form Content Generation92%RANKED19x

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.