💰LLM Pricing War Heats Up: GPT-5.6 Luna Cuts 80%
OpenAI slashes Luna costs, Anthropic follows suit
TL;DR
OpenAI slashes GPT-5.6 Luna costs by 80%, now at $0.02 / $1.20 / $0.20. Anthropic counters with a permanent price cut for Claude Sonnet 5. This pricing war could shift the economics of LLM usage.
OpenAI just cut the price of GPT-5.6 Luna by 80% to $0.02 / $1.20 / $0.20. Anthropic responded by making their temporary price cut for Claude Sonnet 5 permanent. This pricing war is heating up as both companies try to undercut each other. Developers using these models for large-scale projects will see significant cost savings, but the real winners are those using Luna for extensive, low-cost operations. The cache read cost, which accounts for 96% to 99% of total token expenditure, has been slashed, leading to a 70-75% reduction in overall model usage costs. This could drastically change the economics of deploying large language models in production.
Key Points
OpenAI cuts GPT-5.6 Luna prices to $0.02 / $1.20 / $0.20, effective July 30, 2026.
Anthropic makes permanent price cut for Claude Sonnet 5 to $2 / $10 / $0.20.
GPT-5.6 Luna's cache read cost is 96-99% of total token expenditure.
75% reduction in cache read cost correlates to 70-75% reduction in model usage costs.
DeepSeek Flash V4.1 1017 model performs better than V4, hosted by 29 providers on OpenRouter.
Why It Matters
If you're running large-scale projects with GPT-5.6 Luna, the 80% price cut slashes costs significantly. For instance, a project that previously cost $1000 now costs $200. However, the real impact is on those using Luna for extensive, low-cost operations, where the 70-75% reduction in overall model usage costs makes Luna a game-changer. But remember, the economics shift dramatically above ~100K read IOPS, where the premium pricing starts to make sense.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,567 builders reading daily.