OpenAI Releases GPT-6.1 Sol: Near-Astra Coding and Computer Use at One-Fifth of Astra’s Token Price
MarkTechPost Michal Sutter ● Covered by 16 sources
OpenAI shipped GPT-6.1 Sol, its mid-tier GPT-6 model, in the API and ChatGPT Work. It claims near-Astra coding and computer use at one-fifth the token price.
Based on reporting by MarkTechPost, Michal Sutter — 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 has released GPT-6.1 Sol, a new version of its mid-tier GPT-6 model. The pitch is simple and aggressive: close to Astra on agentic coding, computer use, and professional work, but far cheaper to run. OpenAI says the standard input and output prices come in at one-fifth of Astra’s.
The pricing is where Sol makes its case. Input is $2 per million tokens, output is $10, and cached input drops to $0.10. That cached rate matters for agents, which keep sending the same system prompt, tool schema, and history over and over. OpenAI says cached reads now cost 5% of the uncached input rate, down from 10% on GPT-6 Sol.
It is already live as gpt-6.1-sol in the OpenAI API, and also available in ChatGPT Work and Codex. The model offers a 1,050,000-token context window, up to 128,000 output tokens, and supports text and image input with text output. Developers get five reasoning settings: low, medium, high, xhigh, and max. None and minimal are not supported.
On benchmarks, OpenAI is leaning hard on efficiency. On DeepSWE v1.1, Sol matches Astra at roughly one-fifth the cost and beats GPT-6 Sol’s best score by 6.4 points. On GDP.pdf, it beats Claude Opus 5.5 with fallbacks at less than half the cost per task. On OSWorld 2.0 offline, it lands within 2.1 points of Astra at roughly one-seventh the cost per task. On Terminal-Bench Science 0.1, it more than doubles GPT-6 Sol’s score at max effort, though Astra still leads and OpenAI says Astra is the pick for the hardest research.
There are some practical wrinkles. Prompts above 272,000 input tokens cost more, batch and flex are 50% cheaper, and fast mode costs 2x standard. US and EU data residency are supported, but fast mode is not available with EU residency. Fine-tuning is off the table. OpenAI also says a GPT-6.1 Sol Ultrafast option is coming to Codex within days, with up to 8x faster token generation than standard speed.
My take — AI-written commentary, not fact-checked reporting
This is OpenAI doing what OpenAI does best: making the expensive model look too expensive to ignore, then making the cheaper one good enough to rent by the million tokens. The real story isn’t “near-Astra”; it’s that agents are now priced for constant chatty self-billing, which is either progress or a very neat way to hide the meter. Closed models are still the business here, no matter how shiny the benchmark slide looks.
Read more about this at: MarkTechPost