AI Model Selection Matrix
Pick where you'll use the model (Cursor, OpenAI, Claude…), then the job. Rankings use different Artificial Analysis scores per job — no single model wins everything.
Ranked by multi-step work + reasoning scores (not writing-code alone)
Cursor is subscription-based (~$20–$200/mo + add-ons), not published $/1M — we show context window, plus estimated API rates only to compare relative expense
- Architect / Plan
- Execution
- Research
- General use
- Image & video
Key · Fit
Fit (0–100) is how strong a model is at “Architect / Plan” compared with the other models on this platform and budget. It uses different Artificial Analysis scores per task—so the best planner is not automatically the best builder or the cheapest chat model.
Recommended for Architect / Plan
GPT-5.6 Sol
Fit 98/100 · 1.1M context · AA 58.9/77.4/54 · Est. API $5.00/$30.00 per 1M
- Most capableGPT-5.6 Solfit 98 · 1.1M context · $5.00/$30.00
- Highly capableClaude Fable 5fit 95 · 1M context · $10.00/$50.00
- Strong pickClaude Opus 4.8fit 91 · 1M context · $5.00/$25.00
- BalancedGPT-5.6 Terrafit 88 · 1.1M context · $2.50/$15.00
- Good valueClaude Sonnet 5fit 84 · 1M context · $2.00/$10.00
- Best valueGrok 4.5fit 81 · 500K context · $2.00/$6.00
Showing 31 of 56 scored models · Cursor · updated Jul 21, 2026, 10:50 AM ·