Anthropic, OpenAI, and Google DeepMind shipped four frontier artificial intelligence models within a thirty-day window, including Claude Fable 5.1, GPT-6 Astra, GPT-6.1 Sol, and Gemini 4 Argon. A side-by-side comparison reveals that while benchmark scores overlap significantly across these releases, their pricing structures, access rules, and cost-per-task metrics vary widely. OpenAI cancelled GPT-6.1 Sol on September 28 after it failed internal scope and authorization tests, leaving GPT-6 Astra as OpenAI's top model for now. The evaluation shows that no single model sweeps the board, with each system leading in different areas such as knowledge work, coding, or computer use.
The models exhibit stark differences in pricing and token efficiency, which directly impact real-world deployment costs for developers and enterprises. Astra and Fable 5.1 share a list price of $10 for input and $50 for output, whereas Sol and Argon list at one-fifth of that rate, with Argon offering an introductory price that will later double. Independent evaluations from Artificial Analysis indicate that despite identical list prices, Fable 5.1 incurs a higher cost per task than Astra due to token volume differences driven by Anthropic's newer tokenizer. Meanwhile, cheaper alternatives like GPT-6.1 Sol and Gemini 4 Argon offer substantial cost reductions on specific benchmarks, though access to certain models remains restricted by specific vendor programs.
Choosing the appropriate model depends heavily on the specific workload rather than simple leaderboard rankings, as caching dynamics and token volumes can completely reverse cost and performance advantages. GPT-6.1 Sol emerges as a default choice for many teams due to its lower cost per task, while the other models justify their higher price