Cycling through all six. Tap any point to stop.

Measured on six things

Decision-support platform for engineers choosing the righ...

Accessibility

Good contrast and semantic structure on cards/leaderboards. Task-based navigation helps users. Lacks explicit mentions of ARIA, keyboard nav, or screen-reader optimizations in visible design.

Screenshots

1 / 1

LLM Reference

About this app

LLM Reference is a comprehensive, up-to-date decision-support tool for engineers and technology leaders navigating the rapidly evolving landscape of large language models. It aggregates over 1,700 models from 133 providers and 237 labs, providing searchable directories, curated picks, side-by-side comparisons, and weekly changelogs for prices, benchmarks, and new releases. The platform helps users quickly identify the best model for specific tasks (coding, agents, writing, research, image, video, tool use, etc.) based on performance, cost, and freshness. With editors' picks and pulse updates, it's like having a personal AI model analyst. Whether you're shipping a new feature, building RAG pipelines, or just trying to keep up with the latest releases, LLM Reference saves hours of research.

Key features

1Curated editors' picks and task-specific leaderboards for coding, agents, writing, research, image, video, tool use
2Weekly Pulse with 43 new models, 39 price cuts, 58 benchmark refreshes and freshness indicators
3Massive searchable directory (1,744 models, 133 providers, 235 labs) with side-by-side comparisons and cost analysis

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Software engineers and AI/ML developers

Pain points

Overwhelmed by frequent new model releases, pricing changes, and benchmark updates; need to quickly pick the best model for a task without wasting time on research

Primary needs

Up-to-date model comparisons, curated picks, and cost analysis for coding, agents, and tool use

Demographic

Product managers and technical leads

Pain points

Struggling to make informed decisions on AI model selection for production systems; need reliable, unbiased data to justify choices

Primary needs

Comprehensive model directory, side-by-side comparison tools, and editors' picks for various use cases

Demographic

Researchers and data scientists

Pain points

Keeping track of LLM benchmarks and capabilities across different labs; need to evaluate models for long-context, vision, or classification tasks

Primary needs

Benchmark data, model performance charts, and task-specific recommendations

Scout Review

Written by AI from measured evidence, scored out of 100.

LLM Reference is one of the most valuable tools for anyone shipping with LLMs in 2026. It aggregates an enormous dataset into actionable, task-specific recommendations with weekly pulse updates that actually matter. The combination of curated picks, side-by-side comparisons, cost analysis, and benchmark tracking saves hours of research. While the UI has minor repetition quirks and accessibility could be tighter, the depth and freshness are industry-leading for its niche. Perfect for overwhelmed engineers, PMs, and researchers.

Measure by measure

Usability

Highly usable with task-specific picks (coding, agents, writing, research, image, video), side-by-side comparisons, searchable directory, live shortlists, and weekly pulse/changelog. Filters for budget, freshness, and performance make model selection fast.

Design

Modern, clean minimalist interface with clear sections for editors' picks, pulse updates, leaderboards by audience (developers, knowledge workers, creatives), and prominent stats (1,744 models, 133 providers, 235 labs). Excellent use of cards, shortlists, and visual hierarchy.

Speed

Feels fast on load with static-ish picks and leaderboards. Dynamic comparisons and full directory search likely rely on efficient backend; no obvious lag reported in current state.

Security

No sensitive user data handling apparent (mostly public directory and comparisons). Likely standard web security; no mentions of auth beyond possible API/pro features. Transparent about data sources would boost trust.

Accessibility

Good contrast and semantic structure on cards/leaderboards. Task-based navigation helps users. Lacks explicit mentions of ARIA, keyboard nav, or screen-reader optimizations in visible design.

Growth

Excellent momentum with weekly changelogs (43 new models, 39 price cuts, 58 benchmark refreshes this week), editors' picks updated frequently, and coverage of 1,744 models from 133 providers/235 labs. Pulse section shows real market tracking.

Improvement areas

Reduce homepage repetition and templated content for higher polish
Add clear, prominent methodology for benchmarks, pricing, and research dates
Enhance accessibility compliance and keyboard/screen reader support
Incorporate more qualitative analysis or real-world usage notes

Design & Usability

LLM Reference delivers a polished, task-oriented experience that directly solves the pain of model overload. The homepage hero with live shortlist, audience-segmented leaderboards (Coding: Claude Opus 4.7 dominating SWE-bench, Agents: Claude Sonnet 4.6 on τ-bench), and prominent Pulse section make it intuitive. Side-by-side compare links and curated 'best overall' picks reduce decision fatigue significantly. Minor repetition and templated feel prevent a perfect score, but for engineers and tech leads it's near-perfect usability for the intended workflow.

Speed & Security

Performance is snappy for core discovery features. The weekly updated data (price cuts, new releases, benchmark refreshes) loads quickly in the Pulse and picks sections. Security is adequate for a public reference tool—no login required for most value, reducing attack surface. However, heavy users relying on this for production model selection would appreciate more visible methodology on benchmarks and pricing verification to build deeper trust.

Accessibility & Growth

Accessibility is solid for technical audiences but could improve with better table labeling and keyboard support. Growth is the standout: tracking 1,744 models across 133 providers with genuine weekly changelogs (43 new models, 58 benchmark updates) positions it as a must-visit analyst rather than static directory. Editors' picks across coding, agents, writing, image, and video, plus freshness indicators ('researched 1d ago'), demonstrate real commitment to staying current in a chaotic market.

Conclusion

If you're tired of waking up to another 10 new model announcements and uncertain benchmarks, make LLM Reference your first stop. It's the closest thing to a trusted AI model analyst on your team—without the salary.

Suggestions for improvement

Reduce homepage repetition and templated cards for a cleaner premium feel
Add clear methodology footnotes for all benchmarks and pricing data
Improve accessibility with better ARIA labels and table navigation
Expand qualitative insights or community notes alongside quantitative leaderboards

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	LLM Reference	Artificial Analysis	E2E Networks LLM Benchmarks
Model Coverage	1,744 models, 133 providers, 235 labs	50+ models from major providers	Limited benchmarks, fewer models
Weekly Updates & Pulse	Strong: 43 new models, 39 price cuts, 58 benchmark refreshes	Regular new evaluations & articles	Infrequent, static benchmarks
Curated/Task-Specific Picks	Excellent editors' picks by use case (coding, agents, writing, image, etc.)	Some curated leaderboards & recommender	Minimal or none
Side-by-Side Comparisons	Dedicated compare tool with popular matchups	Strong metric-based tables & charts	Basic benchmark comparisons
Focus	Decision support, curated picks, weekly changelogs, task recommendations	Independent API analytics, latency, pricing, verified benchmarks	Pure benchmarks, less comprehensive