Research
9 Frontier AI Models Tested Live (Nov 2026): Pricing Inversions, Thinking-Token Traps & Reproducible Eval Harness
The Brief
- Between November 9 and 18, 2026, Google, OpenAI, and Anthropic shipped six flagship and mid-tier models in a ten-day window.
- The author ran a controlled live benchmark of nine frontier AI models instead of relying on vendor launch benchmarks.
- The write-up highlights pricing inversions and thinking-token cost traps, and publishes a reproducible eval harness.
Why it matters
Vendor benchmarks are often cherry-picked, so independent, reproducible tests help developers compare models on cost and performance before choosing one.
Who’s affected
- Developers choosing between frontier AI models
- Teams budgeting LLM API costs
Between November 9 and November 18, 2026, Google, OpenAI, and Anthropic shipped six flagship and mid-tier models inside a ten-day window. Every launch post came with its own cherry-picked benchmark crown. Rather than trusting press releases, I ran a controlled live benchmark on…