> ## Content Index > Fetch the complete content index at: https://www.blankidentity.com.au/llms.txt > Use this file to discover other available public pages before exploring further. # GPT-6 or Claude Opus 5.5? A plain-English guide to September's new AI models - URL: https://www.blankidentity.com.au/gpt-6-claude-opus-5-5-explainer/ - Published: 2026-09-27T08:09:03.000Z - Updated: 2026-09-27T08:09:03.000Z - Description: OpenAI's GPT-6 Astra, Sol and Luna and Anthropic's Claude Opus 5.5 all arrived in September. Here is what each model does, which plans unlock them in Australia and how to pick the right one. - Author: Matt J - Tags: TECH, Feature September 2026 has been the busiest month for AI models in a long time. OpenAI launched its GPT-6 generation, starting with the flagship GPT-6 Astra earlier in the month and following up with GPT-6 Sol and GPT-6 Luna on 22 September. The same day, roughly 90 minutes earlier, Anthropic released Claude Opus 5.5\. Both companies are promising more capability for less money, and both are rolling the new models out to Australian subscribers now. Here is what each model is, who can use it, what it costs and how to choose. ## The OpenAI line-up OpenAI now has three GPT-6 models, arranged by capability and cost. ![GPT-6 Astra announcement graphic](https://cdn.sanity.io/images/wy75wyma/production/c5ddfbccf448318840ac58e65ba69d34c5321ab5-2240x1260.png) GPT-6 Astra is OpenAI's top-of-range model. Image: OpenAI **GPT-6 Astra** is the top of the range. OpenAI calls it "the world's most intelligent and aligned model" and pitches it at computer use, software engineering, cybersecurity, science and professional work. It is the most expensive of the three through the API, at US$10 per million input tokens and US$50 per million output tokens, with a Fast mode that runs up to twice as fast at twice the price. **GPT-6 Sol** is the everyday workhorse for complex tasks such as coding. OpenAI says it "can take on difficult work tasks while giving you more room to iterate with higher usage limits and lower cost", and TechCrunch reports the company claims it makes about half as many mistakes as its predecessor. It costs US$2 per million input tokens and US$10 per million output tokens. **GPT-6 Luna** is the lightweight model, designed for clerical work such as summarising documents and extracting information. At US$0.10 per million input tokens and US$0.50 per million output tokens, it is by far the cheapest. Sol and Luna cost half as much as the GPT-5.6 models they replace. ![GPT-6 Sol and Luna, with a glowing sun and crescent moon against a dark star-filled sky](https://images.ctfassets.net/kftzwdyauwt9/7fvmMcBIvMTXxVJh7NEMt1/48a75e624699bb1e218366c1d3522048/gpt-6-sol-luna-seo.png?w=1600&h=900&fit=fill) GPT-6 Sol and Luna launched on 22 September 2026\. Image: OpenAI ## The Anthropic line-up Anthropic has taken a different approach, upgrading the largest of its three Claude tiers first. **Claude Opus 5.5** is, according to TechCrunch, what Anthropic calls "the strongest-performing model we've tested to date", and the company says it matches or beats its more restricted Fable model across many benchmarks. Anthropic says Opus 5.5 is 30 per cent faster than Opus 5 and around 40 per cent cheaper on typical workloads. API pricing is US$4 per million input tokens and US$20 per million output tokens. It is pitched at agentic coding, knowledge work, financial analysis and code review, and Anthropic says it writes with less jargon and leads with the most important information. The smaller **Sonnet 5.5** and **Haiku 5.5** models are due "in the coming weeks", TechCrunch reports. ## Who gets what in Australia For most people, the question is not API pricing but which app and plan unlocks which model. Based on what the companies have announced: - **ChatGPT Free and Go:** GPT-6 Luna, via the ChatGPT desktop app. - **ChatGPT Plus, Pro, Business, Enterprise and Edu:** GPT-6 Sol and Luna in ChatGPT Work and Codex. OpenAI says Sol and Luna are not yet in the standard Chat interface. - **GPT-6 Astra:** rolling out over the coming days to ChatGPT Plus, Pro, Business and Enterprise users, included within existing subscription allowances with the option to buy credits for more. Pro, Business and Enterprise users also get GPT-6 Astra Pro. Enterprise administrators must switch Astra on, as it is off by default. - **Claude Opus 5.5:** available in Claude's apps and on the Claude Platform, with Anthropic raising five-hour usage limits on its Pro, Max, Team and seat-based Enterprise plans. Neither company has announced Australia-specific pricing changes alongside these launches. API usage for both is billed in US dollars. ## How to choose Benchmarks from both companies are self-reported and designed to make their own models look good, so treat headline numbers with caution. OpenAI, for instance, says Sol beats Anthropic's previous Opus 5 on its AutomationBench test at a fraction of the cost; Anthropic says Opus 5.5 leads on tests including Terminal-Bench 4.0 and scores 81.8 per cent on the OSWorld 2.0 computer use test. Both claims can be true on the tests each company chose. A more practical way to think about it is by job: - **Quick summaries, email triage and pulling details from documents:** a lightweight model such as GPT-6 Luna is built for this, and it is the one free ChatGPT users can reach. - **Coding, spreadsheets and multi-step work tasks:** GPT-6 Sol and Claude Opus 5.5 are both aimed squarely here. If you already pay for one service, the new model on that plan is the natural place to start. - **The hardest research, security and engineering problems:** GPT-6 Astra is OpenAI's answer, while Anthropic positions Opus 5.5 as close to its Fable model at a lower price. For businesses building products on these models, cost per task now matters as much as raw capability. Both companies emphasised price cuts this month, and caching discounts (Anthropic charges US$0.20 per million tokens for cache reads on Opus 5.5) can make a large difference for apps that reuse the same context repeatedly. ## Safety is part of the pitch Both launches came with heavy emphasis on safeguards. Anthropic says Opus 5.5 carries the same cybersecurity and biology safeguards as Fable 5.1, with verified access programs for security professionals and life sciences researchers. The release is its first since chief executive Dario Amodei called for the industry to pace capability gains so safety research can keep up. OpenAI says Astra supports Zero Data Retention for eligible API customers, and it is testing what it calls Private Safety Processing to strengthen safety monitoring while preserving privacy. ## What's next There is more to come within days. OpenAI's DevDay developer conference takes place in San Francisco on 29 September 2026 (US time), and Fortune reports the company has held back releases for two weeks to concentrate them at the event, including a preview of a cybersecurity-focused GPT-6 Cyber model and what Fortune describes as a dozen or more other products. Anthropic's Sonnet 5.5 and Haiku 5.5 are also expected in the coming weeks. For Australian users, the short version is this: if you pay for ChatGPT or Claude, you already have access to a markedly cheaper and more capable model, and it is worth revisiting tasks that the previous generation struggled with. *Sources:* [*OpenAI*](https://openai.com/index/introducing-gpt-6-sol-and-luna/?ref=blankidentity.com.au) *and* [*Anthropic*](https://www.anthropic.com/claude-opus-5-5?ref=blankidentity.com.au)