🔍 Read the full analysis: Fable, Opus 5.5, Astra, Sol, Luna: Analyzing The Best AI Models For Your Money on ThorstenMeyerAI.com
Get everyday essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
This article compares leading AI models—Fable, Opus 5.5, Astra, Sol, and Luna—focusing on cost, performance, and suitability for different tasks. Opus leads in aggregate performance, while Astra offers a lower-cost alternative, and Sol and Luna enable scalable deployment.
Recent benchmark testing of five prominent AI models—Fable, Opus 5.5, Astra, Sol, and Luna—has revealed significant differences in performance and cost-efficiency. While models like Opus and Astra achieve similar aggregate scores, their pricing and operational efficiencies vary considerably, impacting how organizations should select AI solutions for different tasks. This comparison highlights the importance of matching model capabilities with specific application needs rather than relying solely on advertised token prices or reputation.
According to recent data from Thorsten Meyer AI, Opus 5.5 outperforms other models in aggregate performance, leading in six of ten Intelligence Index evaluations. Its benchmark cost per task is approximately $7.63, despite a listed token price of $10 per million input tokens and $50 per million output tokens. Astra, while priced similarly at $10/$50 per million tokens, achieves a lower benchmark cost of $3.26 per task at maximum effort, offering a more cost-effective solution for tasks requiring similar performance levels.
Models like Sol and Luna, priced at a fraction of Astra and Opus, deliver lower aggregate scores—37 and 48 respectively—but enable large-scale deployment at minimal costs, with Luna costing as little as $0.07 per task. Meanwhile, Fable 5.1, despite a premium reputation, is now challenged by Opus’s superior performance at a lower cost, raising questions about its value proposition for demanding knowledge work. These findings suggest organizations should evaluate models based on task-specific performance and cost, rather than list prices alone.
ThorstenMeyerAI.com / Reality Check
Five models.
Which one earns its cost?
Compare capability, effort and the cost of usable work.
Claude Fable 5.1 · Claude Opus 5.5 · GPT-6 Astra · GPT-6 Sol · GPT-6 Luna
01 Model choice and effort belong together
Anthropic entries include default fallback. Effort labels do not standardize compute across vendors.
| Model | Max effort | Medium effort | Input / output per 1M tokens | ||
|---|---|---|---|---|---|
| Score | Cost / task | Score | Cost / task | ||
| Fable 5.1 | 53 | $7.63 | 49 | $2.98 | $10 / $50 |
| Opus 5.5 | 58 | $5.98 | 51 | $1.34 | $4 / $20 |
| GPT-6 Astra | 53 | $3.26 | 50 | $1.54 | $10 / $50 |
| GPT-6 Sol | 48 | $1.06 | 40 | $0.25 | $2 / $10 |
| GPT-6 Luna | 37 | $0.07 | 29 | $0.02 | $0.10 / $0.50 |
Scores are not success percentages. Benchmark costs are not production quotes or costs per accepted result. Token rates exclude caching discounts and other charges.
02 A shortlist to test on your work
Editorial evaluation proposals—not benchmark-certified specialties.
Constrained, high-volume tasks
Start with LunaTest extraction, classification and transformations against inexpensive, explicit checks.
Recurring development and operations
Trial SolMeasure completion quality and escalation frequency on routine work.
Demanding professional workflows
Compare Opus + AstraTest deliverables, tool execution and review time. Include medium effort before defaulting to max.
Where Fable fits: keep it where a demonstrated task advantage or an established workflow justifies its premium. Require a replacement to earn the switch.
Measure cost per accepted result
Model + tools + review + rework spendingdivided by accepted results. Keep completion time and error severity alongside it.
Sources: Artificial Analysis model pages linked in the table; effort-setting pages below. Figures checked 23 September 2026. The 57% comparison is calculated as 1 − $3.26 / $7.63, rounded. Values may change.
Effort-setting sources and editorial context
Implications for AI Procurement Strategies
The comparison underscores that choosing an AI model is a complex decision involving more than token prices. Opus 5.5’s leading aggregate performance makes it suitable for high-stakes, knowledge-intensive tasks, while Astra’s lower costs make it attractive for routine or high-volume applications. Sol and Luna, with their low costs, are ideal for large-scale deployment where task complexity is moderate. For organizations, this means that a one-size-fits-all approach is ineffective; instead, strategic evaluation based on specific task requirements, expected reasoning depth, and operational context is essential to maximize ROI and efficiency.
Products Worth Considering
DJI Osmo Action 4 Essential Combo Waterproof 4K Action Camera
DJI Osmo Mobile 8 3-Axis Phone Gimbal Stabilizer
Apple AirPods Pro (1st Generation) with MagSafe Charging Case
As an affiliate, we earn on qualifying purchases.
Evolving AI Model Landscape and Benchmarking Methods
The current evaluation builds on a series of recent benchmarks that compare AI models across multiple dimensions, including aggregate intelligence scores, task-specific performance, and operational costs. Notably, the release of Opus 5.5 marked a significant step forward, with leading scores in analytical quality and reasoning. Astra’s emphasis on application-specific capabilities reflects a trend toward models optimized for particular domains such as scientific research and software engineering. Meanwhile, models like Sol and Luna demonstrate the importance of scalable deployment at minimal cost, especially for organizations with large data processing needs.
Prior to this, models like Fable and earlier GPT versions were often evaluated based on their reputation and token prices, but recent data shows that actual performance and operational efficiency are more decisive factors. The benchmarking methodology now considers not only raw scores but also the cost per task at maximum effort, providing a more realistic view of value for money. As AI continues to evolve, ongoing testing and real-world validation remain critical to inform procurement decisions.
“Organizations should evaluate AI models based on specific task requirements rather than just token prices or reputation.”
— Thorsten Meyer
Uncertainties in Model Performance and Cost Comparisons
While the latest benchmark data provides valuable insights, some uncertainties remain. The exact performance of models like Luna and Sol in real-world, domain-specific tasks has not been fully validated beyond initial testing. Additionally, the impact of different interface implementations, software integrations, and operational environments on overall efficiency is still being studied. The comparison also relies on maximum effort settings, which may not reflect typical usage scenarios for all organizations. Further testing and longitudinal data are needed to confirm these findings across diverse application contexts.
Next Steps in AI Model Evaluation and Adoption
Organizations should conduct pilot tests of shortlisted models—particularly Opus 5.5 and Astra—in their specific workflows to validate performance and cost savings. Vendors are expected to release updates and new versions that could shift the competitive landscape. Continued benchmarking and real-world case studies will inform best practices for deploying AI at scale. Additionally, as AI capabilities expand, integration with existing systems and user training will become increasingly important to realize the full benefits of these models.
Key Questions
Which AI model offers the best value for high-complexity tasks?
Based on recent benchmarks, Opus 5.5 leads in aggregate performance and is recommended for complex knowledge work, despite its higher operational costs.
Can Astra be a cost-effective alternative to Opus?
Yes, Astra achieves similar performance at a lower benchmark cost, making it suitable for application-heavy tasks where cost efficiency is a priority.
Are lower-cost models like Sol and Luna suitable for enterprise use?
Sol and Luna are best suited for large-scale deployment where moderate performance suffices, due to their significantly lower costs but lower aggregate scores.
What factors should organizations consider beyond token prices?
Organizations should evaluate models based on performance, task-specific suitability, operational efficiency, and integration capabilities to maximize value.
Will benchmark results remain stable over time?
Benchmark scores can change with model updates and new releases; ongoing testing is necessary to keep procurement aligned with current capabilities.
Source: ThorstenMeyerAI.com
