AI Price War: Claude Opus 5.5 vs GPT-6 Sol and Luna Explained
On September 22, 2026, Anthropic released Claude Opus 5.5 with a price cut of roughly 40% over Opus 5. Within hours, OpenAI answered with two new models — GPT-6 Sol and GPT-6 Luna — both priced at roughly half of what their predecessors cost. It was the clearest sign yet that the fight between the two biggest AI labs has shifted from raw capability to raw cost per token.
⚡ Quick facts
- Claude Opus 5.5: $4 / million input tokens, $20 / million output — cached reads down 60% to $0.20
- GPT-6 Sol: $2 / million input tokens, $10 / million output — half of GPT-5.6 pricing
- GPT-6 Luna: $0.10 / million input tokens, $0.50 / million output — OpenAI's new budget tier
- Timing: Both companies' announcements landed within hours of each other on September 22, 2026
- Why it matters: Lower token prices directly cut the cost of running AI products at scale
What actually changed in pricing?
Claude Opus 5.5 now costs $4 per million input tokens and $20 per million output tokens. Anthropic says that works out to about 40% cheaper than Opus 5 on a typical workload, with the steepest cut going to cached reads — down 60% to just $0.20 per million tokens. For developers who repeatedly reuse the same context (a common pattern in coding agents and long conversations), that cache discount matters as much as the headline rate.
OpenAI's response landed the same day. GPT-6 Sol, a new mid-tier model slotting under the flagship GPT-6 Astra, launched at $2 per million input tokens and $10 per million output — exactly half of what GPT-5.6 cost. Alongside it, OpenAI introduced GPT-6 Luna, a budget-tier model priced at just $0.10 per million input tokens and $0.50 per million output, aimed at high-volume tasks where every fraction of a cent adds up.
Why a same-day price war?
Model quality between the top labs has converged enough that price and latency now decide which API a developer picks for any given task. When Anthropic cuts input pricing on its flagship model, it makes Opus 5.5 more attractive for complex reasoning workloads that were previously too expensive to run at scale. OpenAI's counter with two new tiers — rather than one price cut — covers both ends: Sol competes directly on mid-tier reasoning tasks, while Luna undercuts everyone on cost for simple, high-volume jobs like classification or short summarization.
Think of it like two airlines matching each other's fares within hours of a schedule change — neither wants to be seen as the more expensive option while the other holds a live promotion.
What this means if you build with AI APIs
For developers and companies running AI features at scale, this round of cuts translates into lower operating costs without needing to switch providers or wait for a new model generation. Workloads that route intelligently between models — using a cheap tier like Luna for simple calls and a stronger tier like Opus 5.5 or Sol for harder reasoning — stand to benefit the most, since the price gap between "good enough" and "best available" just widened in the budget model's favor.
It's also a signal that neither lab expects this to be the last cut. Token pricing has fallen sharply through 2026 as both companies compete for developer mindshare, and today's rates are unlikely to be the floor.
Frequently asked questions
How much does Claude Opus 5.5 cost?
Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens — about 40% cheaper than Opus 5 on a typical workload. Cached reads dropped 60% to just $0.20 per million tokens.
What is GPT-6 Sol and how much does it cost?
GPT-6 Sol is OpenAI's new mid-tier model, priced at $2 per million input tokens and $10 per million output tokens — half the price of its GPT-5.6 predecessor.
What is GPT-6 Luna?
GPT-6 Luna is OpenAI's budget-tier model, priced at $0.10 per million input tokens and $0.50 per million output tokens, aimed at high-volume, low-complexity tasks.
Why did Anthropic and OpenAI cut prices on the same day?
On September 22, 2026, Anthropic released Claude Opus 5.5 with a price cut, and within hours OpenAI released two new GPT-6 models at roughly half the cost of their predecessors. The near-simultaneous timing reflects intensifying competition between the two labs for developers building AI-powered products, where token price directly affects margins.
Related articles
Claude Opus 5.5 Explained: Anthropic's New Flagship Reasoning Model
Read more →OpenAI's Model Misalignment Reporting Framework Explained
Read more →Sources: Anthropic and OpenAI official pricing announcements, September 22, 2026 (2026).