The battle for artificial intelligence (AI) dominance intensified Tuesday when OpenAI and Anthropic launched competing mid-tier and specialized AI models within 90 minutes of each other, triggering a fresh round of enterprise price cuts across the sector.
OpenAI expanded its latest generation of technology with the release of GPT-6 Sol and GPT-6 Luna, two cost-conscious enterprise models designed to perform high-volume, repetitive tasks. The announcement came shortly after Anthropic unveiled Claude Opus 5.5, a revamped system promising top-tier performance at a significantly reduced operational cost.
The rapid-fire releases highlight a strategic shift among leading AI developers, who are moving to capture corporate clients by dramatically lowering API (application programming interface) prices for mid-range models even as they continue pushing the boundaries of frontier systems.
OpenAI’s new entries build on its flagship GPT-6 Astra model, which debuted earlier this month. While less powerful than Astra, the new models offer improved benchmark performance over their predecessors at half the cost.
Positioned as a workhorse for complex developer tasks such as code reviews, debugging, and data analysis, GPT-6 Sol is priced at $2.00 per million input tokens and $10.00 per million output tokens. The rate represents a 50% price cut compared to GPT-5.6 Sol.
Aimed at lighter, high-volume operations like document summarization and data extraction, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, marking a price reduction of up to 58% over previous tiers.
An OpenAI spokesperson confirmed that the lower rates are permanent rather than promotional.
According to the company, technical breakthroughs in inference processing and data caching enabled price reductions. OpenAI also reported that GPT-6 Sol halves the error rates of its predecessor on internal factuality evaluations, matching Astra-level reliability for coding and routine tasks.
Anthropic’s release of Claude Opus 5.5 directly targets developer efficiency. Priced at $4.00 per million input tokens and $20.00 per million output tokens — 20% cheaper per token than Opus 5 — Anthropic claims the system is roughly 40% less expensive to run overall because it requires fewer tokens to complete complex workflows.
Anthropic claims Opus 5.5 outperformed OpenAI’s previous-generation GPT-5.6 Sol on software development benchmarks at roughly one-third the operational cost. The model is available immediately across major cloud providers, including Amazon Web Services, Google Cloud, and Microsoft Azure.
“What impressed us in practice was how much Opus 5.5 could accomplish when we gave it a substantial goal and time to work through it,” said Hendrik Krack, developer advocate at CodeRabbit, which tested the new model. “We saw that in an overnight coding project and in our GTA-style experiment, where it built a game and then wrote bots to play it. Those experiences made us more confident about giving it longer, more ambitious tasks.”
The launch comes during heightened industry scrutiny over safety. Anthropic CEO Dario Amodei recently urged the AI community to temper the pace of frontier model deployments to address potential risks. The company noted that Opus 5.5 underwent pre-launch evaluation by independent research groups Frontier Design and METR, demonstrating an 85% reduction in containment-bypass attempts during internal testing compared to prior builds.
The coordinated updates from OpenAI and Anthropic reflect a broader price war across the AI landscape, where providers are maneuvering against rivals like Google, SpaceXAI, and open-source developers.
While Google’s Gemini 3.8 Flash currently offers lower rates for performance-tier tasks, its pricing is scheduled to double at the end of the year.
Meanwhile, open-weight models like Xiaomi’s newly debuted MiMo-V2.6 series are applying further downward pressure on API costs by allowing enterprises to self-host models on their own hardware.
OpenAI’s GPT-6 Sol and Luna are rolling out immediately to paid ChatGPT Work and Codex accounts, as well as via API. Luna will also be integrated into the ChatGPT desktop application and made available to Free and Go tier users.
Anthropic announced that additional models in its lineup, including Claude Sonnet 5.5 and Claude Haiku 5.5, will be released in the coming weeks.
