Key points
- Two top AI labs cut model prices on the same day
- Anthropic says its cheaper Opus nears its flagship
- Both labs tout performance gains alongside lower prices
Anthropic and OpenAI released lower-priced AI models hours apart on Tuesday, each claiming better performance for the money. Anthropic priced Claude Opus 5.5 20% below its predecessor, while OpenAI introduced GPT-6 Sol and GPT-6 Luna at roughly half their predecessors' prices.
Anthropic says Opus 5.5 matches its flagship, Claude Fable 5.1, on most tasks. OpenAI reported competitive results against Anthropic's models on selected tests at lower cost. Both companies cited their own published evaluations.
What each model costs
Anthropic set Claude Opus 5.5's API price at $4 per million input tokens and $20 per million output tokens, down from $5 and $25 on Claude Opus 5, a cut of 20%. Cache reads fall to $0.20 per million from $0.50, a drop of 60%.
Anthropic separately estimates that typical workloads cost about 40% less on Opus 5.5 than on Opus 5. That measures the cost of completing work rather than the per-token discount alone. The company also says output generation is more than 30% faster. Anthropic said cheaper models lower in its lineup, Claude Sonnet 5.5 and Claude Haiku 5.5, will follow in the coming weeks.
OpenAI priced GPT-6 Sol's API at $2 per million input tokens and $10 per million output, and GPT-6 Luna at $0.10 and $0.50. Each is about half the price of its GPT-5.6 version. Sol is aimed at coding, data analysis, and multi-step reasoning. Luna is aimed at high-volume routine work, such as summarizing documents and answering quick questions. Both carry a context window of about 1.05 million tokens.
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Claude Opus 5.5 (Anthropic) | $4 | $20 |
| GPT-6 Sol (OpenAI) | $2 | $10 |
| GPT-6 Luna (OpenAI) | $0.10 | $0.50 |
The three models sit at different tiers, so the prices are not a like-for-like comparison. Opus 5.5 is a high-end model, Sol a mid-tier one, and Luna the cheapest of the group. Each model's listed token prices are lower than its predecessor's.
What the benchmarks show
On Anthropic's published table, Opus 5.5 scores higher than OpenAI's flagship GPT-6 Astra on several coding and reasoning benchmarks, including FrontierCode, where Anthropic lists Opus 5.5 at 54.4% against 53.3% for Astra. On the same table, Astra scores higher on business-workflow automation and on scientific-research tasks. Anthropic compiled the comparison and notes that the GPT-6 figures are the ones OpenAI reported.
OpenAI says Sol made about half as many factual errors as its predecessor on an internal evaluation, and reported competitive results against Anthropic's models on selected coding and business-workflow tests, at lower cost. The comparisons combine company-run evaluations and previously published results, and model versions and testing settings vary, which limits direct comparisons. OpenAI describes that mixture in its methodology note.
Opus 5.5 is also Anthropic's first model since chief executive Dario Amodei published an essay urging the industry to slow the pace of frontier development. Anthropic says Opus 5.5 underwent extensive internal safety testing and was evaluated before release by outside organizations including METR, part of the third-party review both labs have publicized this year.
Neither company is publicly traded. Their cloud partnerships connect the launches to Amazon (AMZN), Alphabet (GOOGL), and Microsoft (MSFT): Claude Opus 5.5 is available on Amazon Web Services, Google Cloud, and Microsoft Azure, and OpenAI's models run on Microsoft's cloud.
