Model-selection guides
Start with the workload, then ask what is true right now.
These guides explain the criteria that actually change a model recommendation. ModelShortlist then lets your AI assistant apply them to current Artificial Analysis benchmark evidence and current OpenRouter operational facts.
Start here
Understand the model-selection framework before narrowing to a specific workload.
How to choose an AI model
Separate hard constraints from preferences, then weigh benchmark quality, economics, and current operational fit.
Read guideWhy ModelShortlist exists
Why the useful question is “best for this workload right now,” not “which model ranks first forever.”
Read guideHow recommendations stay current
See how launches, pricing, benchmarks, capabilities, providers, and ZDR availability can change the answer.
Read guideWorkloads
Start from the job being done and the failure modes that would make a model unusable.
Models for coding agents
Choose models for autonomous and long-running coding workflows using tools, context, agentic evidence, and cost.
Read guideModels for document extraction
Balance extraction reliability, structured output, context, benchmark evidence, and high-volume token economics.
Read guideModels for structured output
Compare models for strict JSON, schema-constrained generation, and tool-driven machine-readable workflows.
Read guideLarge-context models
Compare models for repositories, long documents, transcripts, and workloads where context is a hard constraint.
Read guideCost, tools, and privacy constraints
Apply non-negotiable operational constraints before reasoning about softer quality preferences.
Cost-efficient tool-calling models
Find tool-capable models when repeated agent calls make reliability and token economics equally important.
Read guideModels under $10 / 1M output tokens
Apply a hard output-price ceiling, then compare the strongest eligible models using current evidence.
Read guideZDR model selection
Treat Zero Data Retention as an explicit endpoint-level requirement without restricting ordinary requests.
Read guideEvidence and comparisons
Understand what each data source or comparison method can and cannot tell you.
Artificial Analysis in ModelShortlist
How independent benchmark and performance evidence enters the shortlist with conservative identity matching.
Read guideArtificial Analysis model comparison
Use independent benchmark evidence with current operational facts instead of treating a ranking as the entire deployment decision.
Read guideOpenRouter model comparison
Compare current OpenRouter context, capabilities, pricing, providers, and optional ZDR facts by workload.
Read guideModelShortlist vs static leaderboards
Why benchmark tables are valuable evidence but not a complete workload-specific deployment decision.
Read guideModelShortlist vs model routers
Recommendation and runtime inference routing solve different problems; see where ModelShortlist fits.
Read guideWant the current answer for your workload?
Install ModelShortlist and ask the model-selection question in normal language inside your MCP-capable assistant.