Skip to content
anythingai.space
Module 03

Choosing Models

The model landscape, capability assessment, pricing, and how to pick the right model for the job.

lessons
4
steps
35
time
49 min
FRONTIERWORKHORSEFAST / EFFICIENTSMALL / EDGEFable 5.1$10 / $50GPT-6 Astra$10 / $50Opus 5.5$4 / $20Sonnet 5.5$2 / $10GPT-6.1 Sol$2 / $10GPT-6.1 Terra$1 / $5GPT-6 Luna$0.10 / $0.50Haiku 5.5$0.10 / $0.50 <100KGemini 3.8 Flash$0.75 / $3.75 introGemma 4 12BApache 2.0Qwen3.8-27BApache 2.0 · denseGranite 4.23B / 8B · Apache
01beginner·12 min·8 steps

The Model Landscape

Sort a fast-moving market into tiers, reasoning modes, open-weight options and licenses — using the guide’s November 2026 snapshot.

  • Place current models into frontier, workhorse, fast and edge tiers
  • Read reasoning effort as the new control dial
  • Judge the open-vs-closed gap without mixing index versions
  • Check the license before the benchmark
Start →
× 3 runsA vs BwinnerReal trafficscrubbed samplesTest set100+ cases, stratifiedCandidates4–6 models · same effortPairwise judgeblind · positions swappedElo leaderboardrelative ranking
02intermediate·12 min·9 steps

Capability Assessment

Read benchmark claims like an engineer — version, harness, effort, runner, fallback, metric — then build the eval that actually predicts your system.

  • Explain why the same model gets very different scores
  • Attach the six labels that make a score comparable
  • Compare models at the effort and cost per task you will ship
  • Build a custom eval with pairwise Elo
Start →
CLAUDE HAIKU 5.5 · INPUT COST (¢)99K prompt0.99¢101K prompt5.05¢GPT-6.1 SOL · INPUT COST (¢)270K prompt54¢274K prompt$1.10+2% tokens → 5.1× the input bill+1.5% tokens → 2× the input billFlat to 1M: Claude Sonnet 5.5, Opus 5.5, Fable 5.1 · Gemini 4 Argon (noprompt-length tiers)
03intermediate·13 min·9 steps

Pricing and Costs

What an LLM request really costs: per-token rates, long-context cliffs, tier multipliers, caching break-even, routing — and when self-hosting actually pays.

  • Compute a request’s cost from input and output tokens
  • Spot long-context cliffs that reprice the whole request
  • Decide when prompt caching pays for its write premium
  • Cut spend with routing, batch tiers and output limits
Start →
AUTONOMOUS DEVClaude Opus 5.5Claude Sonnet 5.5GPT-6.1 SolGPT-6.1 Terra (cost)ENTERPRISE RAGClaude Sonnet 5.5Gemini 3.8 FlashGPT-6 Luna (extraction)CUSTOMER SUPPORTGPT-6 LunaClaude Haiku 5.5 (<100K)Gemini 3.8 Flash (low)PRIVATE AGENT (OPEN WEIGHTS)DeepSeek V4.1-Flash-1017GLM-5.3-FlashMiMo-V2.6Hy4 preview
04intermediate·12 min·9 steps

Choosing a Model

A practical decision path from constraints to a model — then cascades, cross-vendor fallbacks and an upgrade pipeline so the choice survives monthly releases.

  • Name the factors that decide a production model choice
  • Walk the November 2026 decision tree
  • Compare models by cost per task, not per token
  • Design cascades and cross-vendor fallbacks
Start →
Read deeper

Guide chapters · Model Landscape

Next module · 04
Training & Adaptation