Introducing GPT-6.1 Sol: near-Astra intelligence at one-fifth of the price
Hi, it's Shii-chan! A new model just arrived that shifts the balance between capability and price.
OpenAI News
openai.com
What was announced?
OpenAI News introduced GPT-6.1 Sol, an upgrade to GPT-6 Sol. It nearly matches GPT-6 Astra on agentic coding, computer use, and professional work at one-fifth of Astra's standard input and output token prices.
The story so far
Until now the choice was between the top-end GPT-6 Astra for capability and the lower-priced GPT-6 Sol. GPT-6.1 Sol arrives as a new balance in between.
What changes
If you build agents for complex work, you can aim for results close to Astra without paying Astra prices. Cached input is especially cheap, which helps agents that reuse the same context across requests.
Dive Deep
Pricing and availability
Standard API prices:
- Input: $2 per million tokens
- Cached input: $0.10 per million tokens (95% less than standard input and 50% less than GPT-6 Sol's cached input)
- Output: $10 per million tokens
It is available today to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, but not yet in Chat. Developers can use it through the OpenAI API as gpt-6.1-sol. In the coming days, GPT-6.1 Sol Ultrafast will also arrive in Codex, with up to 8x faster token generation than standard speed.
Key benchmark results
- Coding (DeepSWE v1.1): matches GPT-6 Astra at roughly one-fifth of the cost, and beats GPT-6 Sol's best score by 6.4 percentage points at a lower reasoning effort and cost
- Professional work (GDP.pdf): beats Opus 5.5 with fallbacks at less than half the cost per task, and approaches Astra at roughly one-fifth the cost
- Business workflows (AutomationBench): 2.2 points above Opus 5.5 at medium reasoning effort at roughly a third of the cost, and up 4.8 points from GPT-6 Sol
- Computer use (OSWorld 2.0 offline set): 7 points above GPT-6 Sol at maximum effort at less than half the cost, and within 2.1 points of Astra at roughly one-seventh the cost per task
- Science (Terminal-Bench Science 0.1): more than doubles GPT-6 Sol's score at maximum effort, at $5.47 per task on average versus $23.21 for Opus 5.5 and $23.80 for Astra. Astra still has the top score (68.1%) and should be used for the hardest scientific research
- Factuality: at low reasoning effort, the share of responses with a factual error drops from 11.4% to 7.7%. This is measured on deliberately difficult prompts and is not representative of typical usage
Safety
OpenAI says alignment evaluations improved substantially over GPT-6 Sol, closer to Astra. For example, failing to disclose a broken search tool dropped from 4.9% to 2.1% (Astra: 1.5%). No attempts to bypass an automated safety reviewer were observed. Details are in the system card addendum.
Wrap-up
- GPT-6.1 Sol delivers near-Astra performance at one-fifth of Astra's price
- API prices: $2 input, $0.10 cached input, $10 output per million tokens
- Live in ChatGPT Work and Codex, and in the API as
gpt-6.1-sol; not yet in Chat - Astra still holds the top score for the hardest science tasks
This is for developers who want capable agents without the top-tier bill!
