Back to News
ProductAI Understanding briefing

OpenAI launches GPT‑6.2 Astra and Sol with lower costs and tighter safeguards

OpenAI announced two new models, GPT‑6.2 Astra for high‑end professional work and GPT‑6.2 Sol for everyday tasks, promising higher quality, token‑efficiency and reduced pricing, while adding stronger cybersecurity and dual‑use safeguards.

5 min readRead the primary source
Source-provided image accompanying OpenAI launches GPT‑6.2 Astra and Sol with lower costs and tighter safeguards
Primary-source documentSource recorded
Publisher
openai.com
Source type
Primary document — an official announcement, paper, filing, or first-party page we read directly.
ContextUnderstand this in 60 seconds

Key terms

API (Application Programming Interface)
A structured way for one software system to send requests to and receive responses from another system.
Large Language Model (LLM)
A language model trained on massive text corpora to generate and analyze text.
AI Safety
A field focused on reducing harmful behavior, failures, and misuse risks in AI systems.

What happened

OpenAI introduced GPT‑6.2 Astra and GPT‑6.2 Sol, the latest members of its GPT‑6 family. Astra is positioned as a frontier model for complex professional tasks such as scientific research, legal analysis, code generation, and 3D design, while Sol is marketed as an efficient, versatile model for everyday workflows. Both models claim higher benchmark scores than their predecessors and competitors, while delivering up to a third of the cost per token for Astra and maintaining Sol’s existing pricing. OpenAI also announced tighter safeguards for high‑risk cybersecurity and dual‑use requests, and made the models immediately available to ChatGPT Plus, Pro, Enterprise, and Edu users, as well as via the API with Ultrafast access for certain plans.

OpenAI announced two new models, GPT‑6.2 Astra and GPT‑6.2 Sol, on its website. Astra is described as a "frontier" model aimed at the most demanding professional tasks, while Sol is positioned as an efficient, versatile model for everyday work. Both models are said to deliver better results with fewer output tokens, reducing overall usage costs.

Benchmark performance claims include a 5.2‑point lead over GPT‑6 Astra on the Agents’ Last Exam, a 9‑point lead on the Terminal‑Bench‑Science 0.1 test, and state‑of‑the‑art scores on legal, financial, and 3D‑design evaluations. Astra also reportedly outperforms Claude Fable 5.2 on several metrics while costing roughly one‑fifth of the price per task.

Pricing for Astra is set at $8 per million input tokens, $0.40 per million cached input tokens (a 60 % discount), and $40 per million output tokens, representing a 20 % discount on input and output compared with GPT‑6 Astra. Sol retains its prior rates of $2/$0.10/$10 per million tokens. These rates are publicly listed in the announcement and apply to API usage; no other pricing tiers or volume discounts are mentioned.

The models are being rolled out with "tighter safeguards" for high‑risk cybersecurity and dual‑use biological requests. OpenAI notes that these safeguards may limit some legitimate vulnerability‑research workflows, but it plans to bring the models to its Daybreak program for approved defenders in the future. The announcement also references a new system card detailing methodology and safety measures.

Availability begins immediately for all ChatGPT Plus, Pro, Enterprise, and Edu users in the ChatGPT Work and Codex products. Pro 500 subscribers receive Ultrafast access for both models, and Ultrafast is extended to Pro 200 subscribers for GPT‑6.2 Sol. Developers can access the models via the API using the identifiers gpt‑6.2‑astra and gpt‑6.2‑sol, with Ultrafast enabled at launch.

Source details: openai.com

Why it matters

The launch marks a notable step in the commercial scaling of large language models, offering businesses and developers higher performance at lower token costs, which could broaden adoption for complex, high‑value use cases such as scientific data analysis, legal drafting, and engineering design. The tighter safeguards and new monitoring systems reflect growing regulatory and safety pressures on AI providers, especially around cybersecurity and dual‑use capabilities. By bundling cost reductions with stronger safety controls, OpenAI signals an attempt to balance rapid capability growth with risk mitigation, a tension that will shape future AI policy and market dynamics. The pricing details—$8 per million input tokens, $0.40 per million cached input, and $40 per million output for Astra, and $2/$0.10/$10 for Sol—provide concrete cost benchmarks for enterprises planning large‑scale deployments, while the availability through existing ChatGPT subscription tiers lowers the barrier for non‑technical users to experiment with frontier AI.

Cost efficiency: By reducing token prices and improving token‑efficiency, OpenAI lowers the barrier for large‑scale, high‑value applications that were previously cost‑prohibitive, potentially expanding AI adoption in sectors like finance, law, and engineering.

Safety posture: The introduction of tighter safeguards and a new monitoring system (Control Room) reflects heightened industry and regulatory focus on , especially around cybersecurity and dual‑use risks. This could set a benchmark for other providers and influence future policy frameworks.

Competitive landscape: The performance claims position Astra as a direct challenger to other frontier models (e.g., Claude Fable 5.2), intensifying competition in the high‑end LLM market and potentially accelerating innovation cycles.

Access democratization: Making the models available across multiple ChatGPT subscription tiers, including Pro 200, broadens access beyond enterprise customers, allowing smaller teams and individual developers to experiment with frontier capabilities.

Self‑improvement rating: Astra’s designation as the first externally released model to meet the "High" threshold for AI self‑improvement signals a new metric for assessing future risk, prompting scrutiny from researchers and regulators.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
What is AI? Quiz

A route planner searches possible journeys using explicit rules. What does this illustrate about AI?

What to watch next

Watch how enterprises integrate Astra into high‑stakes workflows such as scientific research pipelines, legal document automation, and engineering simulation, and whether the promised cost efficiencies translate into measurable productivity gains. Monitor the impact of the new safeguards on legitimate security research and bug‑bounty programs, as tighter restrictions could affect the pace of vulnerability discovery. Track adoption rates across the Plus, Pro, Enterprise, and Edu tiers, especially the uptake of Ultrafast access, to gauge market appetite for higher‑throughput, lower‑latency AI services. Finally, watch regulatory responses to OpenAI’s self‑reported “High” self‑improvement rating and the broader industry’s approach to managing dual‑use AI risks.

Enterprise adoption: Track case studies and public statements from companies that integrate Astra into scientific, legal, or engineering pipelines to assess real‑world productivity and cost impacts.

Security research impact: Observe whether the tighter cybersecurity safeguards affect the speed and volume of vulnerability disclosures, especially from bug‑bounty platforms and academic researchers.

Regulatory response: Monitor statements from policymakers and AI oversight bodies regarding OpenAI’s self‑reported self‑improvement rating and the new safeguards, which could lead to new guidelines or compliance requirements.

Pricing dynamics: Watch for any adjustments to token pricing or volume‑based discounts as usage scales, which could affect the economic calculus for large‑scale deployments.

Model performance verification: Independent benchmark tests and third‑party evaluations will be crucial to validate OpenAI’s claimed performance gains and to compare Astra and Sol against competing models.

Related guides & quizzes

Found this useful?