Dongcha claims OpenAI released GPT-6.1 Sol, with performance close to Astra and unchanged pricing
BIBIBI
AT A GLANCE
The report states that OpenAI launched GPT-6.1 Sol, whose test performance was close to Astra's, with standard API pricing the same as GPT-6 Sol.
Article
OpenAI 1 updated in just weeks GPT-6 Sol, with performance approaching Astra
Insight: Beating AI Flash update,GPT-6 Sol was released only 1 weeks ago,OpenAI has now launched GPT-6.1 Sol. The new model focuses on improving coding,Computer Use and complex professional tasks, while pricing has remained largely unchanged.OpenAI testing,GPT-6.1 Sol achieved 1.1 in the DeepSWE 75.2%coding benchmark,GPT-6 percentage points higher than the best result of 6.4 Sol, already on par with Astra.Computer Use benchmark also trailed Astra by only 2.1 percentage points.
Third-party testing produced similarly close results. In Artificial Analysis's latest Intelligence Index,GPT-6.1 Sol Max scored 52 points,GPT-6 Astra scored 53 points, and GPT-6 Sol scored 48 points.API standard pricing remains per million Token input 2 USD and output 10 USD, the same as GPT-6 Sol. Cached input has also fallen from 0.2 USD to 0.1 USD. Compared with Astra's 10 USD input and 50 USD output, its standard pricing is only one-fifth as much.GPT-6.1 Sol is now available on API, Codex, and ChatGPT Work. Plus users and above can all access it, though it has not yet entered regular ChatGPT conversations.
Early community feedback has focused primarily on value for money. Many developers believe it is clearly more worthwhile than GPT-6 Sol for everyday Coding Agent use, though some real-world tests still find Astra more reliable on complex tasks. It now appears to bring most flagship-model capabilities down to pricing suitable for frequent use.
Original link https://m.theblockbeats.info/flash/369650
Key points
01
Dongcha Beating AI claims that OpenAI it GPT-6 Sol was released 1 weeks later GPT-6.1 Sol。
02
The report summarizes GPT-6.1 Sol's key improvement areas as programming,Computer Use and complex professional tasks.
03
Reporting cited OpenAI testing, stating that GPT-6.1 Sol achieved 1.1 in the DeepSWE 75.2%coding benchmark,GPT-6 percentage points higher than the best result of 6.4 Sol and on par with Astra; the Computer Use benchmark trailed Astra by 2.1 percentage points.
04
The report cites Artificial Analysis's latest Intelligence Index as stating that,GPT-6.1 Sol Max scored 52 points,GPT-6 Astra scored 53 points,GPT-6 Sol scored 48 points.
05
The reported API standard price is per million Token input 2 USD, output 10 USD, the same as GPT-6 Sol; cached input decreased from 0.2 USD to 0.1 USD。
06
The reported Astra price is per million Token input 10 USD, output 50 USD, and stated that GPT-6.1 Sol's standard price is one-fifth of that.
07
The report states that GPT-6.1 Sol is available on API、Codex, and ChatGPT Work; users on Plus and above can use it, but it has not yet entered regular ChatGPT conversations.
08
The report states that initial community feedback focused on value for money: many developers believe it is better suited to serve as a daily Coding Agent, while others found in hands-on testing that Astra handles complex tasks more reliably.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
The core claim of this briefing is:GPT-6.1 Sol performed close to GPT-6 Astra in tests such as programming and computer use, but the standard API pricing listed in the report is significantly lower. The tests, pricing, availability, and community feedback in the article all come from that report and should not yet be considered officially or sufficiently independently verified.
Why it matters to readers
If the performance, pricing, and availability described in the report are accurate, developers may be able to obtain a model with performance close to Astra's test results at a lower cost and use it for more frequent programming and professional tasks.
Beginners should focus on two points: test scores reflect performance on specific tests and do not mean that all real-world tasks are equally reliable;API pricing is calculated separately for input and output Token and actual costs also depend on usage.
Risks and unknowns
There is no OpenAI official announcement in the evidence, so it is not possible to independently confirm GPT-6.1 Sol's release, pricing, and availability.
DeepSWE 1.1、Computer Use The methodology, conditions, and complete results of the Artificial Analysis tests were not provided in the evidence.
“Performance approaching Astra” and “on par with Astra” are based only on the listed tests; actual performance across all tasks cannot be confirmed.
The community feedback does not provide developer identities, sample sizes, or specific testing conditions.
Related Developments
Loading event timeline…
The stability, rate limits, and actual API cost of complex professional tasks remain unclear.
Related concepts
Token
This term is not in the glossary yet. Browse related concepts in the glossary.
