OpenAI’s latest mid-tier upgrade delivers near-flagship intelligence at one-fifth the cost of GPT-6 Astra, signaling that the AI battleground has shifted from raw power to cost efficiency.
OpenAI just made a massive play to dominate high-volume AI workflows.
The company officially launched GPT-6.1 Sol, an upgraded mid-tier model designed to handle heavy agentic coding, data analysis, and multi-step business tasks.
The big headline here is the price tag. GPT-6.1 Sol nearly matches OpenAI’s flagship GPT-6 Astra on complex tasks, but it costs roughly one-fifth as much to run.
This release changes the math entirely for developers building autonomous agents.
Running complex tool-calling loops on flagship models usually drains developer budgets in hours. Standard input tokens on Sol cost $2 per million, while outputs run $10 per million. Even better, OpenAI aggressively cut prompt caching rates down to just 10 cents per million tokens, making long-context conversations dramatically cheaper.
Sol essentially runs neck-and-neck with Astra and top rival models like Claude Opus 5.5 on coding benchmarks like DeepSWE and complex document tasks. Yet across multi-step benchmark tests, Sol averaged $5.47 per task compared to Astra’s $23.80.
That is a massive margin when you operate dozens of automated agents around the clock.
This launch highlights a clear shift in the AI industry.
Frontier research labs are beginning to hit similar capability ceilings on routine work, meaning raw performance no longer guarantees a win. The real competition now centers on unit economics- who can deliver near-flagship intelligence at a price enterprise customers can actually sustain.
With GPT-6.1 Sol, OpenAI just set a high bar for everyone else to match.
