// AI

OpenAI Releases GPT-5.6 Sol and ChatGPT Work, Faces Copyright Sanctions Bid

By Lysias · July 11, 2026

Key Takeaways

A Crowded Model Family Aims at Cost and Speed

OpenAI has broadly rolled out the GPT-5.6 lineup, which ZDNET describes as three distinct models built for different budgets and use cases. Sol is positioned as the flagship, Terra is aimed at everyday tasks, and Luna is the most economical option of the trio. According to ZDNET, OpenAI said the family had cleared a government review process only days before release, a detail the outlet frames as part of a broader informal testing arrangement between AI labs and the Center for AI Standards and Innovation, tied to a voluntary agreement laid out in a June executive order from the Trump administration.

ZDNET reports that Sol comes with a new “ultra” setting that coordinates multiple agents to complete tasks faster, and that OpenAI has pitched Sol as delivering strong performance without a corresponding jump in cost. On UC Berkeley’s Agents Last Exam benchmark, ZDNET notes that Sol beat Anthropic’s Fable 5 in both adaptive and medium reasoning categories. Terra and Luna, meanwhile, were said by OpenAI to outperform Fable 5 while costing around one-sixteenth as much, per ZDNET’s account of the release. The outlet adds that across coding, general workplace tasks, and everyday applications, the GPT-5.6 family lands roughly on par with Fable 5, but with a clear edge in speed and price.

Alongside the model launch, OpenAI introduced ChatGPT Work, an enterprise-focused tool designed to automate workplace tasks across a range of applications, as described in ZDNET’s reporting. For businesses evaluating AI vendors, the combination of a cheaper, faster model family and a dedicated workplace product signals that OpenAI is trying to compete on cost efficiency as much as raw capability, at a moment when Anthropic’s own flagship models have drawn comparable attention.

Why the Regulatory Backdrop Matters

ZDNET’s reporting places the GPT-5.6 release within a wider pattern involving government oversight of frontier AI models. The outlet recounts how Anthropic’s Fable 5 and Mythos 5 were pulled from availability shortly after their own release on government advisement, before Mythos 5 was reinstated for certain partners on June 26 and export controls on both models were lifted by the Department of Commerce on June 30. Anthropic then began restoring global access to Fable 5 on July 1, though Mythos 5 remains limited to specific institutions, according to ZDNET.

That sequence of interventions and reversals is relevant context for OpenAI’s own government-cleared release of GPT-5.6, ZDNET suggests, because it points to a working relationship between major labs and federal reviewers that is shaping release timing without amounting to formal regulation. For anyone tracking how AI companies interact with policymakers — a dynamic that also feeds into investor sentiment around AI-linked equities and tokens — this back-and-forth is a signal that government involvement in model rollouts is becoming routine rather than exceptional, even as the rules themselves remain informal.

On pricing, ZDNET notes that Anthropic extended paid-plan access to Fable 5 through July 12, after originally planning to shift paid users to token-based billing on July 7. Anthropic said in a post on X, as quoted by ZDNET, that users can apply “up to 50% of your weekly usage limit on Claude Fable 5,” after which they can continue with usage credits or switch to another model. Separately, ZDNET reports that Anthropic’s Sonnet 5 starts at $2 per million input tokens, rising to $3 per million in September, and is now the default model in Anthropic’s Free and Pro plans while remaining available on Max, Team and Enterprise tiers.

Legal Pressure Over Training Data Intensifies

Even as OpenAI expands its product lineup, ZDNET reports that the company is contending with a sanctions bid brought by The New York Times, the Daily News and other media outlets. The publishers allege that OpenAI has concealed evidence bearing on how ChatGPT was trained using copyrighted news content, according to the report. This adds to a growing list of legal disputes over data sourcing that AI developers face as they scale their products commercially.

Notably, ZDNET’s own article discloses that Ziff Davis, its parent company, filed a separate lawsuit against OpenAI in April 2025, alleging copyright infringement related to training and operating its AI systems. Taken together, these cases underscore that questions about how large language models are trained, and whose content they draw on, remain unresolved even as the underlying products reach broader commercial and enterprise audiences. For readers interested in AI-adjacent crypto and data-rights projects, this ongoing litigation is a reminder that intellectual property exposure is a persistent, unresolved risk factor for major AI vendors, regardless of how quickly their models improve.

Hype Check

Claim: OpenAI’s GPT-5.6 family, led by Sol, delivers better performance than rivals at a fraction of the cost, cementing OpenAI’s competitive edge. Reality: ZDNET reports genuine benchmark wins for Sol over Fable 5 in specific reasoning categories, and cost advantages for Terra and Luna, but also notes the family is merely on par with Fable 5 in coding and general work tasks — while OpenAI simultaneously faces a sanctions bid from major news publishers over alleged training-data concealment. Verdict: Mixed. This is not financial advice.

Source

Researched with AI assistance, fact-checked and edited by a human. Not financial advice.

📬 Get Cryptoverse Weekly

The week’s most important crypto & AI stories, straight to your inbox. No spam, unsubscribe anytime.

Double opt-in — we’ll email you a confirmation link.