GPT-6.1 Sol vs. GPT-6 Astra: Same accuracy at 18% of the cost
The New Stack Jessica Wachtel ● Covered by 2 sources
OpenAI’s GPT-6.1 Sol matched GPT-6 Astra in tests, but cost about one-fifth as much. The catch: Astra was sometimes faster, so the cheaper model has to earn its keep.
Based on reporting by The New Stack, Jessica Wachtel — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI launched GPT-6.1 Sol on September 29, only a week after GPT-6 Sol. The pitch was simple: treat it as an upgrade over GPT-6 Sol, and as a much cheaper stand-in for Astra. OpenAI says it comes close to Astra on agentic coding, computer use, and professional work, while also cutting factual errors at low reasoning effort.
The pricing is where the story gets loud. GPT-6.1 Sol is listed at $2 per million input tokens and $10 per million output tokens, with cached input at $0.10. Astra sits at $10 per million input tokens and $50 per million output tokens, with cached input at $1. In OpenAI’s framing, that makes GPT-6.1 Sol a fifth of Astra’s price for normal tokens and a tenth for cached input.
The article’s tests put both models through the same three tasks: CI triage, incident logs, and a resolver spec. In CI triage, both models got every call right on all five runs. Astra was quicker, averaging 24 seconds versus 31 seconds for GPT-6.1 Sol, but GPT-6.1 Sol cost about 2 cents a run compared with 11 cents for Astra.
The incident-log test was the bigger one, with 113,966 input tokens per run. Here, both models again answered everything correctly every time. GPT-6.1 Sol finished in 2 minutes and 20 seconds on average, while Astra took 2 minutes and 53 seconds. Cost was $0.31 per run for GPT-6.1 Sol and $1.57 for Astra.
The resolver spec showed the clearest gap. Both models passed all 120 hidden tests on every run, but GPT-6.1 Sol averaged 7 minutes and 7 seconds, while Astra took 10 minutes and 6 seconds. Astra also generated more output tokens, and the cost difference widened: $0.20 for GPT-6.1 Sol against $1.28 for Astra. Across all 15 runs, GPT-6.1 Sol took 49 minutes and 54 seconds and cost $2.66; Astra took 1 hour, 6 minutes, and 55 seconds and cost $14.77.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI pricing that keeps biting the expensive models: if the cheap one is good enough, procurement stops listening to the brochure. Astra may still have a slight edge elsewhere, but for this kind of work paying five times more looks less like quality and more like nostalgia.
Read more about this at: The New Stack