OpenAI released its GPT-5.6 family of models on Thursday, introducing three distinct variants — Sol, Terra, and Luna — in a bid to reclaim the narrative in an increasingly competitive AI market. The launch follows similar announcements this week from SpaceXAI and Meta, but OpenAI’s messaging is squarely aimed at its primary rival, Anthropic.
The new lineup is designed to span a wide range of use cases, from enterprise productivity to scientific research. CEO Sam Altman told CNBC that Sol is 54% more token-efficient on AI coding tasks compared to previous versions, a claim the company is leaning into heavily as it positions GPT-5.6 as a superior option for developers and engineering teams.
Also read: OpenAI Shuts Down Atlas Browser, Moves AI Features to ChatGPT and Chrome
Three tiers, three price points
OpenAI is segmenting its offering more explicitly than before. Sol, described as the workhorse of the family, is priced at $5 per million input tokens and $30 per million output tokens. Terra sits in the middle at $2.50 input and $15 output, while Luna — the budget option — costs $1 input and $6 output. The company says all three models are now available across ChatGPT, Codex, and the OpenAI API.
This tiered approach mirrors what competitors have done, but OpenAI is betting that its performance benchmarks will justify the pricing. The company cites the Artificial Analysis Coding Agent Index, a benchmarking metric, to claim that Sol sets a new state of the art at 80 points — 2.8 points above Anthropic’s recently released Fable 5 — while using less than half the output tokens, taking less than half the time, and costing about one-third less. OpenAI also claims that Terra performs just above Fable 5, and that Luna outperforms Anthropic’s Opus 4.8.
Also read: Lovable in talks to raise $300M at $13.2B valuation, doubling its worth in six months
Cybersecurity and enterprise focus
OpenAI is positioning GPT-5.6 as its strongest cybersecurity model yet, achieving what it calls frontier performance with significantly fewer tokens. The model supports defensive activities including threat modeling, code review and patching, and blue teaming — simulating attacks on one’s own systems to identify vulnerabilities before real hackers do.
The cybersecurity angle is notable given that the Trump administration previously sought to restrict the model’s rollout over concerns about potential misuse. OpenAI has emphasized the defensive applications in its marketing, likely to preempt further regulatory scrutiny.
Alongside the model family, OpenAI also released ChatGPT Work, a workplace companion tool designed for enterprise teams. Running on desktop, web, and mobile, it handles daily clerical tasks such as drafting documents, spreadsheets, and presentations. The move signals OpenAI’s intent to deepen its enterprise footprint, a segment where Anthropic has gained ground by focusing fixedly on business customers.
Competitive environment heats up
Anthropic has managed to position itself as the likable underdog in the AI race, winning growing enterprise support. OpenAI’s latest benchmarks and pricing structure appear designed to undercut that momentum. By offering a model family that spans budget to premium, OpenAI is trying to make the case that it can serve both cost-conscious startups and high-performance enterprise clients more effectively than Anthropic.
The timing is also important: this week alone, SpaceXAI and Meta have released competing models, underscoring how quickly the field is moving. OpenAI’s GPT-5.6 launch is as much about defending its market position as it is about advancing capabilities.
Frequently Asked Questions
What are the three variants of GPT-5.6?
GPT-5.6 comes in three variants: Sol (the most powerful workhorse model), Terra (an intermediate option), and Luna (a budget-friendly choice). Each is priced differently to suit various use cases.
How does GPT-5.6 compare to Anthropic’s models?
OpenAI claims Sol outperforms Anthropic’s Fable 5 on the Artificial Analysis Coding Agent Index, scoring 80 versus Fable’s 77.2, while using less than half the output tokens, taking less time, and costing about one-third less.
What cybersecurity features does GPT-5.6 offer?
OpenAI calls GPT-5.6 its strongest cybersecurity model yet, supporting defensive activities such as threat modeling, code review and patching, and blue teaming (simulating attacks to find vulnerabilities).
Is GPT-5.6 available now and how much does it cost?
Yes, GPT-5.6 is available across ChatGPT, Codex, and the OpenAI API. Pricing per million tokens: Sol $5 input / $30 output, Terra $2.50 input / $15 output, Luna $1 input / $6 output.
Be the first to comment