OpenAI launches GPT-6.1 Sol with near-Astra performance at one-fifth the cost
OpenAI released GPT-6.1 Sol, a new AI model that approaches GPT-6 Astra performance in agentic coding, computer use, and professional tasks at approximately one-fifth the cost. Priced at $2 per million input tokens and $10 per million output tokens, the model features a 95% discount on cached input. Benchmarks show it scores one point below Astra in the Intelligence Index and achieves 75% on DeepSWE at one-fifth Astra's cost, though it underperforms GPT-6 Sol on a writing benchmark by 153 Elo.
Reference imageEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
GPT-6.1 Sol scores 153 Elo lower than GPT-6 Sol on writing benchmark, analysis finds
An analysis from What's AI reports that OpenAI's GPT-6.1 Sol model underperforms its predecessor GPT-6 Sol on a writing benchmark, scoring 153 Elo lower in its best setting (xhigh). The analysis also finds GPT-6.1 Sol is neither cheaper nor faster, costing $0.13 and taking 5.8 minutes per script compared to GPT-6 Sol's $0.11 and 2.9 minutes on max settings. However, the analysis notes that GPT-6.1 Sol's xhigh setting ties GPT-6 Astra on max (1976 vs 1968 Elo) at roughly one-sixth the price ($0.13 vs $0.85). The source recommends users on Astra switch to GPT-6.1 Sol for writing, but advises those on GPT-6 Sol to keep the older model.
Read sourceOpenAI launches GPT-6.1 Sol, scoring near Astra intelligence at a quarter of the cost
OpenAI has released GPT-6.1 Sol, a new AI model that replaces GPT-6 Sol just seven days after its predecessor's launch. According to the announcement, GPT-6.1 Sol scores only one point below GPT-6 Astra in the Intelligence Index while costing less than a quarter of Astra's cost per task ($0.72 vs $3.26). The model maintains the same pricing as GPT-6 Sol at $2 per million input tokens and $10 per million output tokens, but offers an increased cache read discount from 90% to 95%, resulting in a slightly lower blended price for agentic workloads. Key performance gains include a 4-point increase in the Intelligence Index over GPT-6 Sol, significant improvements in agentic knowledge work benchmarks (AA-Briefcase v1.1 and GDPval-AA v2.1), a 12-point jump in Terminal-Bench 4.0, and an 8-point jump in AA-Omniscience Accuracy with hallucination rates dropping from 60% to 54%. The model also shows a 3-point gain in the Coding Agent Index over GPT-6 Sol. The post attributes these achievements to OpenAI and CEO Sam Altman.
Read sourceOpenAI Devs releases GPT-6.1 Sol with agentic coding, near-Astra performance, 95% discount on cached input
OpenAI Devs has announced the release of GPT-6.1 Sol, an upgraded AI model featuring stronger agentic coding and computer use capabilities. The model offers near-Astra performance and provides cached input at a 95% discount to standard input pricing. GPT-6.1 Sol is specifically designed for complex codebase refactors, deep codebase investigations, and long-running agents across applications. The announcement was made via the official OpenAI Devs account on X, highlighting the model's focus on advanced software development tasks.
Read sourceShow 12 older updatesHide older updates
OpenAI Releases GPT-6.1 Sol with Enhanced Agent Coding and Computer Use
OpenAI has announced the release of GPT-6.1 Sol, a new model that upgrades agent coding and computer use capabilities, approaching the performance of the Astra model. The model features a 95% discount on cached input pricing compared to standard input prices. According to the official announcement from OpenAI Developers, GPT-6.1 Sol is designed for complex refactoring, deep codebase investigation, and long-running agents that operate across multiple applications. The pricing change for cached inputs is specifically highlighted as a factor that can be used to evaluate the model's suitability for long-duration tasks. This release signals OpenAI's continued focus on improving autonomous coding agents and their ability to interact with computer systems over extended periods.
Read sourceOpenAI says GPT-6.1 Sol improves on GPT-6 Sol in complex professional tasks
OpenAI has announced that its new model, GPT-6.1 Sol, delivers significant improvements over its predecessor, GPT-6 Sol, across a range of complex professional tasks. According to the company, the enhancements are particularly notable in areas such as code writing and debugging, document understanding, and executing multi-step business workflows. The announcement was made via a post on X, linking to further details from TechCrunch. This development signals a continued rapid iteration in OpenAI's model lineup, focusing on practical, high-level enterprise and developer use cases.
Read sourceGPT-6.1 Sol scores 75% on DeepSWE at one-fifth the cost of Astra, new benchmarks show
A post from user pankajkumar_dev on X reports benchmark results for a new AI model, GPT-6.1 Sol. On the DeepSWE benchmark, GPT-6.1 Sol achieved 75% at roughly one-fifth the cost of Astra, compared to GPT-6 Sol's 68.8% at $2.74 per task. On AutomationBench, GPT-6.1 Sol medium scored 31.7% at about one-third the cost of Opus 5.5, while Sonnet 5.5 reached 44.7%. On Terminal-Bench Science, GPT-6.1 Sol scored 57% at $5.47 per task, versus GPT-6 Astra at 68.1% for $23.80 and Opus 5.5 at 58.7% for $23.21. The post notes that compared to GPT-6 Sol, GPT-6.1 Sol shows a big jump on DeepSWE and Terminal-Bench Science, with the main story being much better cost efficiency. The author concludes that while GPT-6.1 Sol is not beating Opus 5.5 on every benchmark, its price-to-performance ratio looks very interesting.
Read sourceOpenAI launches GPT-6.1 Sol, claims it nearly matches Astra at 20% token price
OpenAI has launched GPT-6.1 Sol, a new AI model, according to a post on X. The company claims that GPT-6.1 Sol nearly matches the intelligence of Astra for tasks including coding, computer use, and professional work, while being priced at only 20% of Astra's token price. This announcement positions GPT-6.1 Sol as a more cost-effective alternative to Astra for high-performance AI applications. The post does not provide details on availability, benchmarks, or the specific capabilities of Astra being referenced.
Read sourceOpenAI launches GPT-6.1 Sol on OpenRouter at $2/M input, nearly matching GPT-6 Astra at one-fifth price
OpenAI has released GPT-6.1 Sol, now available on the OpenRouter platform. The new model nearly matches the performance of GPT-6 Astra on agentic coding, computer use, and professional work, but at one-fifth of Astra's price. Pricing is set at $2 per million input tokens and $10 per million output tokens, with cached input priced at $0.10 per million tokens, half the cached price of GPT-6 Sol. The announcement was made via a post on X (formerly Twitter) by OpenRouter.
Read sourceOpenAI launches GPT-6.1 Sol with near-Astra intelligence at one-fifth the price
OpenAI has launched GPT-6.1 Sol, a new AI model that offers near GPT-6 Astra level intelligence at a fifth of the price. According to the announcement, across all reasoning settings, the model's error rate stays within 1.9% of GPT-6 Astra. The model also improves factual accuracy when given difficult prompts, with its largest gain over GPT-6 Sol appearing at low reasoning effort, where the share of responses containing a factual error drops from 11.4% to 7.7%. The post also notes that OpenAI is not launching GPT-6.1 Astra, as was originally expected.
Read sourceGPT-6.1 Sol is renamed GPT-6 Astra Minor after botched launch, source says
According to a post on X, OpenAI's GPT-6.1 Sol is merely a renamed version of GPT-6 Astra Minor, following what the source describes as a botched launch of GPT-6 Sol. The source claims the company had to replace the model a week after its release, calling the situation embarrassing. The post further asserts that the renamed model still does not reach the performance level of Astra, let alone Opus 5.5. The information is attributed to an anonymous source on the platform synthwavedd, and no official confirmation or additional details are provided.
Read sourceOpenAI releases GPT-6.1 Sol, claiming near-Astra intelligence at one-fifth the price
OpenAI has announced the release of GPT-6.1 Sol, a new AI model that it claims delivers performance close to its Astra-level intelligence at a significantly reduced cost. The company states that the model offers 'near-Astra intelligence for a fifth of the price,' positioning it as the most cost-efficient option for its performance tier currently available on the market. The announcement was made via a post on X, which included a link to further details. The model appears aimed at providing high-level AI capabilities at a more accessible price point, potentially expanding the user base for advanced AI tools. No specific benchmarks or technical specifications were provided in the announcement, but the claim of cost efficiency relative to performance is the central selling point.
Read sourceOpenAI Releases GPT-6.1 Sol, Nearing GPT-6 Astra Performance at One-Fifth Cost
OpenAI has announced the release of GPT-6.1 Sol, a new AI model that approaches the intelligence of its flagship GPT-6 Astra in agentic coding, computer use, and professional work tasks. The company published benchmark comparisons against GPT-6 Astra and detailed API pricing. The standard API price for GPT-6.1 Sol is $2 per million input tokens, $0.10 for cached input, and $10 per million output tokens, approximately one-fifth the cost of GPT-6 Astra. Developers can now evaluate whether the lower-cost model can serve as a viable substitute for the flagship model in production environments.
Read sourceGPT-6.1 Sol AI model released for long-running agents at $2 input and $10 output pricing
A developer on X announced the release of GPT-6.1 Sol, a new AI model designed specifically for long-running agents. The model is priced identically to its predecessor GPT-6 Sol, at $2 per million input tokens and $10 per million output tokens. The poster expressed curiosity about whether GPT-6.1 Sol will be comparable to Opus 5.5, and stated they are waiting for benchmark results to make that determination. The announcement was made by user pankajkumar_dev on the X platform, linking to further details. No official benchmarks or performance data were provided in the post itself.
Read sourceOpenAI releases GPT-6.1 Sol with 1.05M context window at $2 per million input tokens
A new AI model called GPT-6.1 Sol has been released, according to a post on X by user DanDr1s. The model is priced at $2 per million input tokens and $10 per million output tokens, matching the pricing of the previous GPT-6 Sol. It features a 1.05 million token context window and a maximum output of 128,000 tokens. The model is described as having the highest reasoning tier, supporting up to 'max' reasoning effort. Its knowledge cutoff date is April 30, 2026. The post states the model is built for coding, computer use, and professional work. The announcement includes a link to an external source, but no official confirmation from OpenAI is provided in the post itself.
Read sourceOpenAI DevDay 2026 Unveils GPT-6.1 Sol, Persistent Agent Dots, and Over 20 Updates
OpenAI announced over 20 updates at its DevDay 2026 event, including the release of GPT-6.1 Sol, a new model that demonstrates strong performance in agentic coding, computer use, and professional tasks. The pricing for GPT-6.1 Sol is approximately one-fifth of the standard input/output token price of GPT-6 Astra. Other announcements include the introduction of a persistent agent named Dots and expanded platform openness. The official summary provides a comprehensive overview of the entire product line released at the event.