Fresh off the wire, still hot.
Anthropic just released Claude Opus 5.5 today and cut its price by 20%.
Minutes later, OpenAI clapped back with two new GPT-6 models, Sol and Luna, priced at half of their predecessors.
Think of it like two neighboring gas stations. One drops prices in the morning. The other rolls out a bigger discount by lunch, just to make sure nobody drives past without stopping.
The numbers tell the real story. Opus 5.5 now costs $4 per million input tokens and $20 per million output tokens, roughly 40% cheaper than Opus 5 for typical workloads.
Cache reads got the biggest cut of all: down 60% to just 20 cents per million. That is basically the model reusing what it already 'remembers' instead of thinking from scratch, and now it is dirt cheap.
There is also a faster serving mode for anyone who needs speed over savings, at a premium of $8 per million.
This is not just an intelligence race anymore. It is a race to see who can hand you the same brainpower for less money.
Great news if you are building anything on top of these models, because the cost ceiling that used to hold projects back keeps dropping every few months.
So is this an AI arms race, or a price war wearing an AI costume? Honestly, it is both.
🔗 Original Source & Reference: https://siliconangle.com/2026/09/22/anthropic-releases-claude-opus-5-5-and-openai-counters-with-two-cheaper-gpt-6-models/
Published automatically via FeedMind AI Content Pipeline.
Top comments (0)