GPT-6.1 Sol is OpenAI’s newest coding model, launched at DevDay 2026 on September 29, 2026. It costs the same as its predecessor but performs closer to OpenAI’s flagship model. Codex, OpenAI’s coding tool, also changed. It gained a cloud version, a faster Ultrafast tier, and new GitHub security scanning at the same event. For developers, the surrounding tools matter as much as the model itself. Codex now runs remotely, adds an always-on agent mode, and prepares automated fixes for security flaws.
What Is GPT-6.1 Sol
GPT-6.1 Sol is OpenAI’s updated coding and reasoning model, released September 29, 2026, at DevDay 2026. It replaces GPT-6 Sol as the default coding model in ChatGPT and Codex. Developers can call it directly through the OpenAI API as gpt-6.1-sol. OpenAI built Sol for agentic coding and computer-use tasks. Those are tasks an autonomous agent performs without step-by-step human guidance.
The headline claim is intelligence, not price. Sol keeps the same rate as GPT-6 Sol: $2 per million input tokens and $10 per million output tokens. OpenAI says that pricing sits near one-fifth of GPT-6 Astra’s standard rate, OpenAI’s larger flagship model. Cached input tokens cost $0.10 per million, a 95 percent discount off standard input pricing. That discount rewards agents that reread the same repository or context window repeatedly.
GPT-6.1 Sol Benchmark Performance
OpenAI published benchmark comparisons against its own earlier models and against Anthropic’s Claude Opus 5.5. On DeepSWE v1.1, a coding benchmark, Sol scores 6.4 points higher than GPT-6 Sol. It matches GPT-6 Astra’s performance on that test at one-fifth the cost. On OSWorld 2.0, a computer-use benchmark, Sol beats GPT-6 Sol by 7 percentage points. It lands within 2.1 points of Astra at one-seventh the cost.
On AutomationBench, Sol scores 2.2 points above Claude Opus 5.5, at roughly one-third the cost per task. On Terminal-Bench Science 0.1, OpenAI reports an average cost of $5.47 per task, against $23.21 for Opus 5.5. OpenAI also reported a lower factuality error rate, down from 11.4 percent to 7.7 percent, a 32 percent reduction. These are OpenAI’s own published numbers, not independently audited figures. Treat them as a starting point for evaluation, not a final verdict.
Codex in the Cloud and the CLI Refresh
OpenAI rebuilt Codex, its coding agent, around the same DevDay announcements. Codex in the Cloud lets a developer start a task on one device and finish it on another. It runs in reusable development environments, so a team can share one approved configuration and permission set across a project. That helps teams standardize how AI tools fit into their software development practices. Each engineer no longer configures an agent differently.
Codex CLI, the command-line version, also got a refresh. It now supports voice-controlled task steering, so a developer can direct a task without typing. A new agents view lets a developer track and delegate multiple running tasks at once. Both updates ship across all Codex plans at no added cost.
Codex Security Cloud and the Ultrafast Tier
DevDay 2026 also introduced Codex Security Cloud, a scanning tool built into Codex. It scans an entire GitHub repository on demand or on a schedule. When it finds a vulnerability, it investigates the issue and prepares a fix for review. OpenAI’s own DevDay 2026 recap covers eligibility. It lists the tool as available to Pro, Business, Enterprise, and Edu accounts at no added cost.
OpenAI also launched Ultrafast, a faster processing tier. It generates up to 300 tokens per second. That is up to 8 times faster than standard speed inside Codex, and 6 times faster through the API. GPT-6 Astra Ultrafast is available now. It costs $60 per million input tokens and $300 per million output tokens, six times the standard rate. Sol’s Ultrafast option is not live yet. OpenAI lists it as coming soon.
Choosing the Right Model and Tooling for Your Team
Teams already running GPT-6 Sol in production can switch to gpt-6.1-sol without a pricing change. The bigger decision is which tool changes how the team works. Codex in the Cloud and Security Cloud matter more than which model answers prompts. A team doing routine bug fixes and pull request generation fits standard Sol well. A team running longer, autonomous agent sessions fits a different profile. DevX has covered how generative AI reshapes how software gets built. That team gets more value once the Ultrafast tier reaches Sol.
Security-conscious teams should test Codex Security Cloud against their existing scanning tools before relying on it alone. Automated fix suggestions still need a human reviewer before merging. Treat the tool as a first pass, not a replacement for code review.
Common Mistakes and Tradeoffs to Watch
The most common mistake is assuming GPT-6.1 Sol replaces GPT-6 Astra for every task. Sol closes most of the benchmark gap, but Astra still leads on the hardest reasoning tasks. Reserve Astra, or its Ultrafast tier, for work where that final quality gap matters.
The second mistake is turning on Codex in the Cloud without reviewing what permissions each environment carries. A misconfigured shared environment can expose more of a codebase than intended. Set permissions deliberately before a team adopts shared cloud environments, not after an incident.
Key Takeaways
- GPT-6.1 Sol launched September 29, 2026, at OpenAI DevDay 2026, priced at $2 and $10 per million input and output tokens.
- Sol claims near-GPT-6 Astra intelligence at about one-fifth Astra’s standard price.
- Codex in the Cloud lets developers move a coding task between devices using reusable development environments.
- Codex Security Cloud scans GitHub repositories and prepares fixes for review, included at no added cost on paid plans.
- The Ultrafast tier reaches 300 tokens per second. It is live for Astra now and coming soon for Sol.
Frequently Asked Questions About GPT-6.1 Sol
What Is GPT-6.1 Sol?
GPT-6.1 Sol is OpenAI’s updated coding model, launched September 29, 2026. It targets agentic coding and computer-use tasks at GPT-6 Sol’s original price.
How Much Does GPT-6.1 Sol Cost?
Sol costs $2 per million input tokens and $10 per million output tokens. Cached input tokens cost $0.10 per million, a 95 percent discount.
Is Codex in the Cloud Available Now?
Yes. Codex in the Cloud is live for Codex users on paid ChatGPT plans. It lets developers run tasks remotely with reusable development environments.
Does GPT-6.1 Sol Replace GPT-6 Astra?
No. Sol closes much of the benchmark gap at a lower price. Astra still leads on the hardest reasoning and computer-use tasks.
What Does Codex Security Cloud Do?
It scans GitHub repositories for vulnerabilities on demand or on a schedule. It then prepares a fix for a developer to review and merge.
When Is the Ultrafast Tier Coming to GPT-6.1 Sol?
OpenAI has not given a date. GPT-6 Astra Ultrafast is available now, and OpenAI lists Sol’s Ultrafast option as coming soon.
Final Thoughts
GPT-6.1 Sol matters less as a new model than as a signal of where OpenAI is putting its effort. The price stayed flat while the surrounding tools, cloud environments, security scanning, and faster inference expanded. Before adopting any of it, test Sol against real prompts and confirm what permissions a shared cloud environment actually grants. Keep Astra in reserve for tasks where the last percentage points of accuracy matter. Match the tool to the task, the same rule that applied before DevDay.
Photo by ilgmyzin: Unsplash
Priya Nandakumar covers enterprise technology and AI infrastructure for DevX, with a focus on the systems decisions that look fine until they don't. Caching layers, message queues, fault tolerance. She spent seven years as a backend engineer at two Series C startups before moving into technical journalism, and she still reads changelogs for fun.
