GPT-6.1 Sol Is Insane! + OpenAI's DevDay: Dots, Pro 500, Ultrafast, Codex & More! AI News

The big picture: DevDay became an agent operating system launch

OpenAI’s DevDay 2026 centered on a clear strategic message: the company no longer wants ChatGPT to be only a place where users ask questions; it wants ChatGPT, Codex and the API to become an always-on work layer for agents, teams and developers. OpenAI described the event as its biggest DevDay yet, with more than 20 major announcements across ChatGPT, Codex, models and new AI work surfaces [1]. The common thread was not just “better models,” but cheaper reasoning, persistent agents, faster execution and enterprise controls.

That is why GPT-6.1 Sol matters. It is not simply another model in a crowded picker. OpenAI positions it as a major upgrade over GPT-6 Sol for agentic coding, computer use and professional work, with “near-Astra” intelligence at one-fifth of GPT-6 Astra’s standard input and output token prices [1]. In practical terms, OpenAI is trying to move high-level reasoning from “special occasion” usage into daily software, operations and document workflows.

GPT-6.1 Sol: near-Astra capability at a Sol price

The central claim around GPT-6.1 Sol is cost-adjusted performance. OpenAI says the model nearly matches GPT-6 Astra on agentic coding, computer use and professional work while costing one-fifth of Astra’s standard input and output token prices [2]. The published API pricing is $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens [2]. Compared with the Astra tier described in OpenAI’s own positioning, that pricing turns Sol into the obvious default for teams that need strong reasoning but cannot justify using the top model on every request.

The cached-token number may be the most important detail for real products. At $0.10 per million cached input tokens, OpenAI says cached input is 95% cheaper than standard input and 50% cheaper than GPT-6 Sol’s cached input pricing [2]. That changes the economics of long-running agents. If an assistant needs to reuse project history, repository structure, customer context or policy documents across many calls, caching can keep the “memory tax” from overwhelming the application budget.

OpenAI’s model page reinforces the same positioning: GPT-6.1 Sol is meant for complex coding, computer use and professional work, with a 1,050,000-token context window and up to 128,000 output tokens [4]. The model supports reasoning efforts from low through max, while excluding none and minimal, which signals that Sol is intended for deliberative tasks rather than lightweight instant chat [4].

Benchmarks point to the real target: agentic work

The benchmark story is less about leaderboard theater and more about where OpenAI expects agents to generate value. On DeepSWE v1.1, OpenAI says GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth the cost and beats GPT-6 Sol’s best score by 6.4 percentage points at lower reasoning effort and cost [2]. For software teams, that is the key comparison: not whether Sol is the absolute smartest model in the family, but whether it is good enough to run more often.

The same pattern appears in business and computer-use tasks. OpenAI says GPT-6.1 Sol scores above Opus 5.5 with fallbacks on GDP.pdf at less than half the cost per task, and on AutomationBench it is 4.8 percentage points above GPT-6 Sol at the same reasoning setting [2]. On OSWorld 2.0’s offline set, GPT-6.1 Sol comes within 2.1 percentage points of Astra at maximum reasoning effort while costing roughly one-seventh as much per task [2]. That combination explains the excitement: the model is tuned for the work agents actually do, such as navigating interfaces, reading dense files, debugging systems and chaining steps together.

Availability: powerful, but not everywhere in Chat yet

There is one important limitation: GPT-6.1 Sol is not simply “the new default ChatGPT chat model.” OpenAI says it is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and developers can access it through the API as gpt-6.1-sol; OpenAI also says it is not yet available in Chat [2]. The Help Center release notes similarly say the rollout begins in ChatGPT Work and Codex, starting with Pro users and expanding to Plus, Business, Enterprise and Edu, with workspace admins able to manage access through permissions and roles [5].

That distinction matters for expectations. GPT-6.1 Sol is being launched first where structured work happens: coding sessions, professional workflows, workspace contexts and API calls. It is a builder-and-operator model before it is a casual conversation model.

Dots: the always-on agent layer

If GPT-6.1 Sol is the efficiency story, Dots is the product story. OpenAI describes dots as always-on agents powered by GPT-6 Astra, with their own cloud computer, the ability to learn from feedback over time, and connections to more than 4,000 apps through plugins [3]. OpenAI says a dot can work toward goals around the clock, communicate through ChatGPT, Slack or Teams, and carry context across channels [3].

This is a much more ambitious framing than a custom bot. Dots are designed to monitor work, prepare drafts, investigate bugs, update analyses, build pull requests or manage content production while the user remains in control [3]. OpenAI says dots begin rolling out to Pro and Business Premium users in eligible markets, while Enterprise, Edu and Healthcare users can try the beta when an admin enables it [1]. The first dot is included in Pro or Business Premium plans at no extra cost, but deeper work still counts against usage allowances when the dot starts or manages tasks in Codex or ChatGPT Work [3].

The control model is central. OpenAI says dots use separate cloud computers, app permissions, custom rules, read-only proactive research for background discovery, and auto-review for actions that could affect accounts or share information [3]. That is necessary because an always-on agent is more useful than a chatbot, but also more sensitive.

Pro 500 and Ultrafast: speed becomes a paid feature

DevDay also introduced a sharper segmentation of speed and quota. OpenAI says the new Pro 500 plan provides the highest usage allowance among Pro plans at 25 times the ChatGPT Plus allowance and includes access to Ultrafast [1]. The Help Center says Pro 500 is a $500-per-month plan and that, among Pro plans, Ultrafast is available only with Pro 500 at launch [5].

Ultrafast itself is a premium speed tier for workloads where latency matters. OpenAI says it offers up to 8x faster token generation in Codex, reaching 300 tokens per second, and up to 6x in the API [1]. At launch, GPT-6 Astra Ultrafast is available in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans, while GPT-6.1 Sol Ultrafast is coming soon [1]. For developers building interactive tools, voice-like experiences or rapid coding loops, the message is clear: intelligence is one axis, but responsiveness is now a monetized axis too.

Codex, Agents API and enterprise plumbing

Codex received some of the most practical DevDay updates. OpenAI announced Codex in the cloud, allowing developers to run tasks from a computer, phone or cloud environment, with reusable development environments that preserve approved settings and dependencies [1]. The refreshed Codex CLI adds voice task steering and an /agents view for delegating and tracking multiple tasks [1]. Code review also moves into the ChatGPT desktop app with summaries, diffs and automatic cloud reviews [1].

On the API side, the Agents API now supports computer use, along with Codex multi-agent capabilities, tool search, tool calling and context compaction inside developer applications [1]. The OpenAI API changelog separately lists GPT-6.1 Sol for the Responses API and Chat Completions, notes multi-agent support in beta, and confirms that tool calling should use the Responses API [4]. This is the infrastructure layer beneath the flashier product announcements.

Safety: cheaper does not mean casual

The safety context is important because OpenAI is putting stronger models into more persistent workflows. The GPT-6.1 Sol system card addendum says OpenAI treats the model as Critical in cybersecurity and High for biological and chemical capability under its Preparedness Framework, and applies the same safeguards stack as GPT-6 Astra [6]. OpenAI also says GPT-6.1 Sol showed improved alignment behavior over GPT-6 Sol in areas such as broken-search transparency, respecting restrictions and avoiding unauthorized outcomes during agentic tasks [2].

AP reported that Altman introduced Dots and other DevDay updates shortly after OpenAI halted the rollout of another advanced model over safety concerns, and that he said the company was investing more in safety, security and monitoring of AI agents [7]. That backdrop makes DevDay feel less like a pure acceleration story and more like a controlled expansion of capability: more autonomy, but also more guardrails, admin controls and monitoring.

Why this launch feels different

GPT-6.1 Sol is “insane” not because it claims to beat every model at every task, but because it changes the price-performance curve where agents live. The model is strong enough for serious coding, computer use and professional workflows, cheap enough to run repeatedly, and optimized for contexts where cached tokens and long memory matter [2]. Around it, OpenAI added Dots for persistent personal and organizational agency, Pro 500 for heavy users, Ultrafast for latency-sensitive work, and Codex/API upgrades for builders.

The result is a DevDay that feels like OpenAI connecting the dots literally and strategically. GPT-6.1 Sol lowers the cost of intelligence; Dots give that intelligence a persistent job; Codex and the Agents API give it tools; Pro 500 and Ultrafast create the premium lane. For developers and enterprises, the next question is no longer whether an AI can answer a prompt. It is how many real tasks it can safely keep moving while nobody is watching the screen.

Sources from the last 72 hours

1. [1]DevDay 2026 RecapSep 29, 2026, 2:00 AM
2. [2]Changelog | OpenAI APISep 29, 2026, 2:00 AM
3. [3]ChatGPT — Release NotesSep 29, 2026, 2:00 AM
4. [4]Addendum to GPT-6 Astra System Card: GPT-6.1 SolSep 29, 2026, 2:00 AM
5. [5]GPT-6.1 SolSep 29, 2026, 2:00 AM
6. [6]Introducing dotsSep 29, 2026, 2:00 AM
7. [7]OpenAI CEO Sam Altman announces agent Dots, GPT-6.1 Sol and other updatesSep 30, 2026, 2:00 AM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.