Editorial illustration for OpenAI’s Sol Matches Astra Performance at a Fraction of the Cost
OpenAI Sol Matches Astra at Fraction of Cost
Four frontier models shipped in four weeks, and the launch posts made each one sound like the new champion. Claude Fable 5.1 arrived September 1. GPT-6 Astra followed two days later.
GPT-6.1 Sol and Gemini 4 Argon landed before September closed out. OpenAI quietly scrapped a fifth model, GPT-6.1 Astra, on September 28 after it failed internal scope and authorization tests, leaving GPT-6 Astra as the company's top model by default rather than design.
Put the four survivors side by side and the picture gets messier than any single announcement let on. Benchmark scores cluster closer together than the marketing implies. Pricing does not.
Astra and Fable 5.1 both list at $10 input and $50 output. Sol and Argon undercut that by a factor of five, though Argon's rate is introductory and set to double. For anyone building agents that resend system prompts and tool histories on every step, the cached-input price matters more than the headline rate, and that's where the real gaps show up.
Astra and Fable 5.1 share the same $10 input and $50 output list price. Sol and Argon list at one-fifth of that. Argon’s price is introductory and doubles later.
Why this matters OpenAI shipping Sol as a cheaper near-match for Astra, then killing GPT-6.1 Astra on September 28, tells us more than the benchmark tables do. When a lab retires its flagship nine days before a cheaper sibling lands at a fifth to a seventh of the cost per task, that's a pricing decision dressed up as a product decision. For developers and founders choosing between Fable 5.1, Argon and Sol, the lesson isn't which model wins on DeepSWE or OSWorld 2.0.
It's that these numbers come from the vendors themselves, measured on their own harnesses, and four launches in 30 days makes side-by-side comparison nearly impossible without someone doing the unglamorous work of normalizing cost-per-task across providers. The 2.2-point edge over Claude Opus 5.5 on AutomationBench is real only until someone runs it independently. We'd treat Sol's efficiency claims as a starting point for procurement conversations, not a verdict.
Watch whether Anthropic or Google respond with their own cost-adjusted benchmarks, because that's where this actually gets decided.
Related Reading
Browse all OpenAI news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source
Related Topic OpenAI Developer: AI Agent "Swarms" Waste Tokens, Create "Coordination Tax
AI Tools & Apps
From Archives OpenAI Adds Subtle Text Watermark to ChatGPT in European Union
From Archives OpenAI adds optional AI watermark for API users, mandatory in EU
From Archives OpenAI consults mathematicians after AI missteps in math research
Research & Benchmarks
From Archives OpenAI Tests Visual Ads on Image Results, Labels Them Clearly
Business & Startups
Common Questions Answered
How does GPT-6.1 Sol's pricing compare to GPT-6 Astra's pricing?
GPT-6.1 Sol costs one-fifth of GPT-6 Astra's price, with Sol listing at $2 input and $10 output compared to Astra's $10 input and $50 output. Despite the significant price difference, Sol matches Astra's performance capabilities, making it a more cost-effective option for developers and founders.
Why did OpenAI cancel GPT-6.1 Astra in September?
OpenAI quietly scrapped GPT-6.1 Astra on September 28 after it failed internal scope and authorization tests. This cancellation left GPT-6 Astra as the company's top model by default rather than through intentional design, occurring just nine days before the cheaper Sol model launched.
What four frontier models were released within four weeks in September?
OpenAI and Google released Claude Fable 5.1 (September 1), GPT-6 Astra (September 3), GPT-6.1 Sol and Gemini 4 Argon (before September ended). However, only three of the four models survived, as GPT-6.1 Astra was discontinued on September 28 due to failing internal testing requirements.
What does OpenAI's decision to release Sol at a fraction of Astra's cost suggest about the company's strategy?
OpenAI's release of Sol as a cheaper near-match for Astra indicates that the pricing decision was strategic rather than purely product-based. When a lab retires its flagship model nine days before launching a cheaper sibling at one-fifth to one-seventh of the cost, it suggests the company is prioritizing cost-effectiveness and market positioning over maintaining a premium flagship offering.
What is the introductory pricing structure for Gemini 4 Argon?
Gemini 4 Argon lists at one-fifth of the price of Astra and Fable 5.1, placing it in the same affordable tier as Sol. Argon's current price is introductory and will double at a later date, meaning users will face price increases once the promotional period ends.
Further Reading
- Introducing GPT-6.1 Sol - OpenAI
- Addendum to GPT-6 Astra System Card: GPT-6.1 Sol - OpenAI
- Introducing Claude Fable 5.1 and Claude Mythos 5.1 - Anthropic
- Claude Fable 5.1, Anthropic's new frontier model is now available in Amazon Bedrock - Amazon Web Services
- GPT-6.1 Sol, GPT-6 Sol and Luna: Pricing, Benchmarks, Tests - ComputingForGeeks
