Prices verified September 7, 2026
Compare AI models on price, context & capability
35 models from 13 providers, normalized so the numbers are actually comparable — and checked against provider documentation rather than copied from launch announcements. Prices moved repeatedly in 2026; this table reflects what they are now.
- Cheapest capable
- $0.22 / 1M
- Largest context
- 1.05M tokens
- Comparisons
- 132
Start with what you're building
Reasoning models
Thinking tokens bill as output. See which model is actually cheapest per finished task.
Coding & agents
Coding Agent Index scores next to real token costs, because efficiency beats a two-point lead.
Open weights
Self-hostable models, licences, and when GPU time beats per-token pricing.
Cost calculator
Enter your token volumes and caching rate to see every model's monthly bill.
The current frontier
Three models sit at the top end, and the gap between them is far smaller than the gap in their prices.
OpenAI
$4.00 / $20.00 / 1M
Frontier coding and long-horizon agents
1.05M context • native reasoning
Anthropic
$10.00 / $50.00 / 1M
Long-horizon agentic engineering
1M context • native reasoning
$2.00 / $12.00 / 1M
Best price-to-reasoning ratio
1.05M context • native reasoning
Most looked-up comparisons
Each one includes the price gap in percentage terms and the monthly bill at three realistic workload sizes.
Why trust these numbers?
Most comparison tables quote launch prices. Providers cut prices mid-cycle — one budget tier fell 80% in a single day in July 2026 — so a launch-day figure can be wrong by a factor of five.
Read the methodology →Primary sources, with dates
Every price comes from the provider's own pricing page or API docs, and each model shows when it was last checked. Conditions like promotional windows and long-context surcharges are recorded, not hidden.
No affiliate links
No sponsored placements and no paid rankings. Where we cannot verify a price from a primary source, we leave the model out rather than publish a guess.