Field notes on LLM routing and cost | Nadir
Practical guides to cutting LLM API costs with intelligent routing and context optimization. Written by the engineers shipping it.
Topics
- Routing & Cascades: Classifiers, verifiers, cascades, and the benchmarks that judge them.
- Context & Compression: Making the prompt smaller before anyone bills you for it.
- Caching: Prompt caches, semantic caches, KV caches, and their taxes.
- Agents: Where agentic loops actually spend the budget.
- Pricing & Models: Rate cards, tokenizers, tiers, and the cliffs between them.
- FinOps & Governance: Budgets, caps, mandates, and who answers for the bill.
- Nadir & Alternatives: How we built it, and the honest comparisons.
Pages on this site
Machine-readable summaries of this site:
llms.txt and
llms-full.txt.