Tech • AI • Robotics • Game
GPT-6.1 Sol Is Insane! + OpenAI's DevDay: Dots, Pro 500, Ultrafast, Codex & More! AI News
9.7/10TL;DR
OpenAI unveiled GPT-6.1 Soul alongside a broad slate of products, positioning the model as near-Astra performance at roughly one-fifth the cost while expanding its agent, coding and enterprise tooling.
KEY POINTS
GPT-6.1 Soul pricing push
GPT-6.1 Soul is priced at $2 per million input tokens and $10 per million output tokens, far below GPT-6 Astra at $10 and $50 respectively. Cached input falls to $0.10 per million tokens, a steep cut aimed at making higher-end reasoning practical for daily use and large-scale deployment.
Performance close to frontier models
OpenAI said the model nearly matches GPT-6 Astra across coding, computer use and professional workflows. Reported benchmark results cited 75% on Deep Sway, up from 68.8% for GPT-6 Soul, and 31.7% on Automation Bench, where it trails Sonnet 5.5 at 44.7% but does so at much lower cost.
Strong price-to-performance case
On Terminal Bench Science, GPT-6.1 Soul reportedly scored 57% at $5.47 per task, versus Astra at 68.1% for $23.80 and Opus 5.5 at 58.7% for $23.21. Another intelligence index placed it at 52, just one point behind Astra, reinforcing the case that the new model competes more on economics than outright top-end capability.
Real-world coding and generation tests
In side-by-side tests, GPT-6.1 Soul completed a 3D keyboard assembly task at about 30 times lower cost than Sonnet 5.5, though with slightly lower quality in that domain. It also generated game-style and interactive web outputs including a Minecraft-like sandbox, a 360-degree product viewer, and an Elden Ring-style action game with combat, animation, ambient audio and UI systems, highlighting the model’s ability to produce polished end-to-end code.
One million token context
The model carries a 1 million token context window, matching leading long-context offerings. Testing suggested its best efficiency may come at medium reasoning settings, where usage costs remain lower while output quality stays competitive.
Dots agent launched
OpenAI also introduced Dots, an always-on agent powered by GPT-6 Astra. It runs on its own cloud computer and browser, connects to more than 4,000 apps, and is designed to work continuously in the background, with memory for user preferences, goals and standards.
Enterprise and workflow automation
Demonstrations showed Dots monitoring customer feedback, preparing PR materials, updating research and supporting sales workflows. One example involved the agent noticing an unpaid invoice, drafting it and preparing it to send after approval, illustrating OpenAI’s push toward proactive software agents rather than prompt-by-prompt assistants.
Subscription changes and faster tiers
The $200 Pro plan returned with lower usage capacity than before, though without the old five-hour limit. OpenAI also introduced a new $500 Pro tier with much higher limits and Ultra Fast mode, which promises up to 300 tokens per second in Codex and as much as six times faster API generation speeds.
Codex security and cloud updates
Codex Security Cloud gained a major upgrade through Daybreak Blue, which can continuously scan GitHub repositories, review new commits, investigate vulnerabilities and prepare fixes even when a machine is offline. Codex cloud environments now let agents keep working with preconfigured dependencies and settings while users are away from their computers.
Decision API and collaborative workspace
A new Decision API powered by GPT-6 Luna was introduced for classification, routing and agent decision-making, positioning it as a lightweight orchestration layer for multi-agent systems. ChatGPT Space was also launched as a collaborative workspace with interactive pages containing charts, images, checklists and dashboards that teams can edit together.
CONCLUSION
The announcements mark a clear strategy: lower the cost of frontier-level AI while expanding the surrounding software stack for agents, coding and enterprise coordination. GPT-6.1 Soul appears central to that push, offering competitive capability at pricing aggressive enough to challenge rivals on everyday adoption.
Ask a question