How is the Cost Per Task (CPT) calculated?

CPT is calculated as the average cost of a standard 1 Million token workload (500k input / 500k output) based on live OpenRouter API pricing, divided by the model's domain capability score.

What is the difference between Free and Paid LLM rankings?

Paid models are ranked by capability-weighted cost-efficiency (ROI), while Free models ($0 cost) are ranked strictly by raw benchmark capability and context window capacity.

How often are LLM prices and benchmark scores updated?

Prices are updated automatically every Monday via GitHub Actions querying the live OpenRouter API. Benchmark scores are evaluated against standard suites (MMLU-Pro, HumanEval, MT-Bench, AgentBench).

LLM Cost Ranker

Best cost-performance AI model by capability & budget — updated weekly.

⚡ 300+ Models Analyzed • 🔄 Updated Every Monday

Ranked by CPT / Intelligence ratio

BEST OVERALL — PAID MODELS

—

CPT Score —

Context Window —

Score —

Sources: OpenRouter API ↗ Artificial Analysis ↗

Cost Per Task (CPT)

—

Top 10 Rankings

Ranked by CPT Score — lower is better

Rank	Model	CPT Score	Context	Score	Source

Benchmark Methodology & FAQ

Understanding LLM Cost Per Task (CPT) & ROI metrics

📊 What is Cost Per Task (CPT)?

Cost Per Task (CPT) measures the true inference cost to execute a standard 1 Million token workload (500k input / 500k output) derived from live OpenRouter API pricing data. Rather than evaluating raw single-token rates, CPT reflects real-world developer costs.

⚡ How is Domain Capability Weighted?

Specialized domains (Thinking, Coding, Agentic, Writing) apply non-linear capability exponents to standard benchmarks (MMLU-Pro, HumanEval, MT-Bench, AgentBench). This ensures top-tier frontier models win high-complexity categories while ultra-budget options remain recommended for basic tasks.

🎁 How are Free Models Evaluated?

For $0 cost models, rankings evaluate a Task Efficiency Index combining raw capability scores with logarithmic context window scaling, ensuring generous context windows are rewarded fairly.

🔄 Automated Weekly Sync

Every Monday, an automated GitHub Action syncs prices across 300+ models directly from OpenRouter, recalculating ROI indexes so you always deploy the most cost-effective LLM.