OpenAI Launches GPT-5.6 Sol, Faces US Government Access Restrictions

OpenAI has officially launched its new flagship AI model, GPT-5.6 Sol, a powerful competitor designed to rival Anthropic’s Claude Mythos series. Despite demonstrating superior performance in key benchmarks, particularly in agentic coding, the rollout of GPT-5.6 Sol is significantly constrained by directives from the US government, limiting its availability to only a select group of partners. OpenAI has voiced strong objections to these restrictions, arguing that such government-controlled access impedes broader innovation and keeps essential tools away from developers, enterprises, and cybersecurity defenders.

This development follows a previous instance where the US government withdrew Anthropic’s Mythos-class model Fable 5 from the market, indicating a pattern of increased government scrutiny and intervention in the deployment of advanced AI models.

Key facts:

Feature	Detail
Model Name	GPT-5.6 Sol
Competitor	Anthropic’s Claude Mythos 5
Key Performance	Outperforms in agentic coding (Terminal-Bench 2.1), strong in cybersecurity
Access	Limited to select partners via API and Codex, directed by US government
Pricing (per 1M tokens)	Input: $5, Output: $30

Performance and Capabilities

GPT-5.6 Sol introduces a new layered naming scheme, featuring “Sol” as the flagship, “Terra” as a mid-tier option matching GPT-5.5 at half the cost, and “Luna” as the budget-friendly choice. It also includes “max” and “ultra” modes for enhanced reasoning and parallel sub-agent task execution, respectively.

In benchmarks, Sol has shown significant advancements. On Terminal-Bench 2.1, Sol scored 88.8 percent, with Sol Ultra reaching 91.9 percent, surpassing Claude Mythos 5’s 88 percent. It also demonstrated improvements in genomics and quantitative biology on GeneBench v1, outperforming GPT-5.5. In cybersecurity, specifically on ExploitBench, Sol matched Mythos Preview’s performance in finding and exploiting security flaws while using considerably fewer tokens. OpenAI positions Sol as its most capable cybersecurity model to date, emphasizing its role as a defender for identifying and fixing vulnerabilities rather than an attacker.

Pricing and Efficiency

OpenAI has announced pricing for GPT-5.6 Sol at $5 per million input tokens and $30 per million output tokens. Terra and Luna offer more economical options. The company has also revamped its prompt caching system, introducing explicit cache breakpoints and a guaranteed minimum lifetime of 30 minutes. Despite the per-token costs, OpenAI suggests that Sol’s efficiency in using fewer tokens to achieve or exceed competitor performance could lead to a lower effective cost per task, potentially countering the trend of increasing AI model prices.

Implications for AI Development and Accessibility

The restricted launch of GPT-5.6 Sol highlights a growing tension between rapid AI innovation and government oversight, especially in critical areas like cybersecurity. OpenAI’s public criticism underscores the industry’s concern that such limitations could stifle progress and prevent valuable tools from reaching a wider audience that could benefit from them. For Indian developers, businesses, and cybersecurity professionals, these global regulatory trends signify potential challenges in accessing cutting-edge AI technologies, necessitating a close watch on international policy developments that impact AI deployment. The debate over controlled access versus widespread availability will likely shape the future trajectory of advanced AI models.

Source: The Decoder, https://the-decoder.com/openais-claude-mythos-competitor-gpt-5-6-sol-launches-under-government-controlled-access-it-calls-unsustainable/