Editorial illustration for OpenAI's GPT-6 Sol Outperforms Claude Opus at One-Tenth the Cost
GPT-6 Sol Beats Claude Opus at 1/10th Cost
OpenAI's GPT-6 Sol Outperforms Claude Opus at One-Tenth the Cost
OpenAI added two models to its GPT-6 lineup on Thursday: Sol and Luna, both slotting in below GPT-6 Astra, which launched earlier this month. Both are live now in the API under the names gpt-6-sol and gpt-6-luna. Neither ships with downloadable weights, so this is an API-only release, not something you can run locally.
The pitch is straightforward. Astra handles the hardest work. Sol is aimed at complex coding and professional tasks at a fraction of Astra's cost.
Luna covers fast, high-volume everyday queries. OpenAI says improvements in caching and inference let it cut API prices for Sol and Luna by 50% compared to GPT-5.6 promotional rates, though Luna's output price actually drops from $1.20 to $0.50 per unit, closer to a 58% cut.
The bigger claim is on performance. OpenAI put Sol up against Claude Opus 5 on a coding and agent benchmark called AutomationBench 1.0.6, along with a separate test called Agents' Last Exam. The company says Sol matches or beats Anthropic's model while costing a small fraction as much per task.
OpenAI team states better caching and inference let it serve these models more cheaply. It is cutting Sol and Luna API prices by 50% against their GPT-5.6 promotional pricing.
Why this matters
The number that jumps out here is the 11x cost gap, not the 6.3-point benchmark lead. Sol beating Claude Opus 5 on AutomationBench matters less than the fact that OpenAI can now offer "good enough" performance at a fraction of the price, which is the actual lever for teams building automation pipelines at scale. If you're running thousands of agent tasks a day, a model that scores slightly worse but costs a tenth as much wins every time on unit economics.
We'd also flag that Astra at low effort still undercuts Sol's cost by staying at 3.9x, so OpenAI is clearly tuning multiple price-performance curves rather than just shipping one flagship and calling it done. For developers, the practical move is to test Sol and Luna against whatever Claude or Astra pipeline you're already running and see where the effort settings actually land for your workload, since "xhigh effort" and "high effort" aren't free abstractions, they carry their own cost curves. Worth watching whether Anthropic answers on price rather than raw score.
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's GPT-6 Models Halve Prices but Show Minimal Performance Gains
From Archives Claude Opus 5.5 Cuts Costs By Using 40% Fewer Tokens
From Archives OpenAI Seeks Mathematicians' Advice After AI Math Results Backfire
Business & Startups
From Archives The Download: AI's self-improvement problem and OpenAI's safety pause
Research & Benchmarks
Common Questions Answered
How does GPT-6 Sol's pricing compare to Claude Opus, and what makes it more cost-effective?
GPT-6 Sol outperforms Claude Opus at one-tenth the cost, representing an 11x cost advantage. OpenAI achieved this pricing through better caching and inference optimizations, cutting Sol and Luna API prices by 50% compared to their GPT-5.6 promotional pricing.
What is the difference between GPT-6 Sol and GPT-6 Luna in terms of use cases?
GPT-6 Sol is specifically aimed at complex coding and professional tasks that require sophisticated reasoning. GPT-6 Luna is designed for fast, high-volume operations where speed and efficiency are prioritized over maximum capability.
Why is GPT-6 Sol available only through the API and not as a downloadable model?
GPT-6 Sol and Luna are API-only releases, meaning they cannot be run locally on your own infrastructure. This API-only approach allows OpenAI to maintain control over the infrastructure and optimize serving costs through their centralized caching and inference systems.
How does GPT-6 Sol's performance on AutomationBench compare to other models, and why does cost matter more than benchmark scores?
GPT-6 Sol beats Claude Opus 5 on AutomationBench with a 6.3-point benchmark lead, but the real value proposition is the cost efficiency. For teams running thousands of agent tasks daily, a model that costs one-tenth as much provides superior unit economics, making it the practical choice for automation pipelines at scale.
Further Reading
- Introducing GPT-6 Sol and Luna - OpenAI
- OpenAI launches GPT-6 Sol and Luna, boasting lower cost - TechCrunch
- Anthropic releases Claude Opus 5.5 and OpenAI counters with two cheaper GPT-6 models - SiliconANGLE
- OpenAI launches faster, cheaper GPT-6 Sol and Luna - TestingCatalog
- GPT-5.6 and GPT-6 Pro in ChatGPT - OpenAI Help Center
