AI pricing split: Top models cost 20,000x more than bottom as context grows 63x
move-size123 · reddit · 2026-08-19
An analysis of OpenRouter data reveals a massive split in AI model pricing. The cheapest output (Mistral Nemo) is $0.03 per million tokens, while the most expensive (o1-pro) is $600. Meta's average ($0.74) is 64x cheaper than OpenAI's ($47.63).
Key findings:
- Provider Averages: OpenAI ($47.63) and Anthropic ($44.79) are significantly higher than Google ($5.58), Mistral ($3.68), and Qwen ($2.86).
- Reasoning Costs: Output tokens are the killer for reasoning models. Qwen3 thinking variants charge 12x more for output than input, which can lead to surprisingly high bills for agents running in loops.
- Context Growth: Average context window has increased 63x over three years (from 10.5K to 662K) while prices have remained roughly flat.
- Free Models: 19 models are actually free on the API (not just trial credits), including NVIDIA Nemotron 3 Ultra and Google Gemma 4.
More from Venture
- Oracle down 12% in a week, nearly 50% off June high as OpenAI exposure spooks market — alvelda · 2026-10-16
- Anthropic IPO valuation expectations drop to ~$1.5T as Q3 revenue growth slows to 45% QoQ — scaling01 · 2026-10-16
- VC Brad Gerstner hypes AI on CNBC while his AI exposure sits at 3-year low — SumitGup · 2026-10-16
- Anthropic pre-IPO futures imply $1.9T+ while investors eye $1.5T entry — firstadopter · 2026-10-16
- Investors temper Anthropic IPO valuation expectations over revenue growth slowdown — Ok_Display_3159 · 2026-10-16
- Don't Let the Agent Grade Its Own Homework: What a Backtest Tool Should Return — QuanTradin · 2026-10-16