LLM Model Guide

Pick the right model.
Stop overpaying.

Not every task needs GPT-4. Most don't. Here's what actually matters: the right model for the job, and what it costs per million tokens.

Section 01
What are you building?
Pick a task. Get 2–3 model recommendations with honest takes and real pricing.
👆Pick a task above to see model recommendations.
🧮 What does 1M tokens actually buy you?
▼
Task ~Tokens each 1M tokens = At Haiku pricing ($5/M out)
Cold email written ~600 ~1,600 emails $0.30 total
Customer support reply ~650 ~1,500 replies $0.33 total
Lead classified ~350 ~2,800 records $0.18 total
Sales call summarized ~1,700 ~600 summaries $0.80 total
Blog post outline ~600 ~1,600 outlines $0.30 total
Full doc analyzed ~8,000 ~125 docs $3.20 total
Input counts too. Summarizing a transcript costs more per output than writing a short email — because you pay for every token in, not just out. Run a high-context task through Sonnet instead of Haiku and your cost multiplies fast.
Section 02
Full Cost Reference
All major models, side by side. Output cost bar is relative to Claude Opus 4.6 ($25/M = 100%).
Model Provider Input / 1M tokens Output / 1M tokens Best For

Want these running for your business?

we build AI automations for GTM teams — the right model, the right workflow, the right cost. Let's talk about what that looks like for you.

Book a Map call →