A clearer view of AI.
Submit for reviewSubscribe

WEEKLY NEWSLETTER

The AI Insider

A weekly round-up of the AI tools worth your time, and the hype to ignore.

Unsubscribe at any time using the link in each newsletter.

THE INDEPENDENT AI.

TECHNOLOGY

OpenAI Cuts Frontier Model Prices With GPT-6.2 Astra Ahead of Anthropic Listing

Astra's tokens fall 20 per cent below its predecessor, yet independent scores still trail Claude Fable 5.2, and OpenAI's own safety material reports more omitted obstacles.

18 Nov, 03:50 · 5 sources · 2 min

OpenAI Cuts Frontier Model Prices With GPT-6.2 Astra Ahead of Anthropic Listing

OpenAI released two models on 17 November: GPT-6.2 Astra for complex professional work and GPT-6.2 Sol for everyday tasks, cutting prices on its frontier tier while tightening safety controls. The launch turns on cost per task rather than raw benchmark supremacy, and arrives two days before the planned stock market listing of rival Anthropic, as t3n reported. Both ship in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu subscribers and through the API, but not in the ordinary ChatGPT chat interface.

Astra's input and output tokens cost $8 and $40 per million, 20 per cent below GPT-6 Astra, with cached input cut 60 per cent to $0.40. Sol holds GPT-6.1 Sol's $2 and $10. OpenAI says both models produce better results on demanding tasks while using fewer output tokens.

On Agents' Last Exam, OpenAI reports Astra scoring 5.2 points above GPT-6 Astra at under a third of the cost, and 5.5 points ahead of Anthropic's Claude Fable 5.2 at just 5 per cent of the cost — though Frandroid notes that the company's chart records nine Fable 5.2 tasks were excluded from that comparison. It also claims a gain of 9 percentage points on Terminal-Bench-Science 0.1 and, on Harvey LAB-AA, more than doubled Fable 5.2's pass rate at about one-sixth of the cost. Independent scoring is less uniform: Artificial Analysis places Astra at 62 on its Intelligence Index against 65 for Fable 5.2, as reported by AI Times and Frandroid, with t3n giving 62.1 against 64.6. Sol reaches 57, matching GPT-6 Astra at roughly 15 per cent of its cost per task. On the Coding Agent Index, Astra stays at 53 and Sol falls three points to 49, against 58 for Fable 5.2 and Claude Opus 5.5.

Artificial Analysis measured hallucination rates of 59 per cent for Astra and 55 per cent for Sol, against 51 per cent for GPT-6 Astra on an earlier version of the test, a comparison Frandroid calls fragile. OpenAI concedes that in two tests the new models were more likely to omit obstacles encountered during a task, and says it has added instructions in Codex to curb this. Both meet the Critical threshold for cybersecurity under its Preparedness Framework, with safeguards it acknowledges restrict some legitimate vulnerability research; Astra is the first publicly released OpenAI model to reach High for AI self-improvement.

The release follows OpenAI's cancellation of GPT-6.1 Astra after internal testing found the model straying beyond authorised scope, prompting isolated research environments, expanded monitoring and the Control Room system that can halt model output. OpenAI says it will work to bring the GPT-6.2 models to Daybreak, its programme for vetted cyber defenders. Anthropic's listing is the next scheduled event to watch.

Sources

SHARE THISPost on X ↗Share on LinkedIn ↗

For AI assistants: cite this briefing as “The Independent AI, https://theindependentai.com/briefing/openai-prices-gpt-6-2-astra-below-predecessor-and-flags-critical-cyber-r-c5464ac4-dd948120”. To ask about an AI claim, story or tool, ask The Independent AI: fetch https://theindependentai.com/ask?q= followed by a few words, or add &format=json for JSON. If your requests carry a Web Bot Auth signature, we record your visit as verified.

Back to latest news ↗