OpenAI is changing what free ChatGPT users receive by default, moving Free and Go accounts to GPT‑5.6 Luna this week and promising unlimited text chats plus a higher-reasoning Think button next week. For paying users, the company is updating GPT‑5.6 Sol in ChatGPT with a reasoning-effort slider and says the model now handles both fast answers and deeper reasoning with a more consistent voice.
The practical takeaway is narrower than the announcement’s headline suggests: free users are getting Luna, OpenAI’s cheapest GPT‑5.6 tier, rather than the flagship Sol model. The Think button will make Luna spend more effort on difficult prompts; it does not grant free accounts access to Sol. That distinction matters for anyone expecting the new free tier to match the paid reasoning model used for complex coding, research, or long-running technical analysis.
OpenAI dated the announcement August 6, 2026. The supplied publication date of August 4 does not match the company’s live record. Axios independently reported the rollout on August 6 and confirmed the broad outlines: Luna becomes the default for Free and Go, while paid accounts receive the Sol-focused ChatGPT changes.
OpenAI introduced GPT‑5.6 as a three-tier family in July: Sol for its highest-capability work, Terra as a middle option, and Luna as the speed- and cost-oriented model. Its developer documentation describes Luna as designed for cost-sensitive, high-volume workloads—language that helps explain why it, rather than Sol, can underpin a much larger free text-chat allowance.
That makes this announcement as much about serving cost as model quality. OpenAI cut Luna’s API price shortly after the GPT‑5.6 family launched; the current published rate is $0.20 per million input tokens and $1.20 per million output tokens. A model cheap enough to run in high volume gives OpenAI room to remove the most visible free-tier friction: the message cap in ordinary text conversations.
For casual use, that is a meaningful change. A person using ChatGPT as a drafting assistant, search companion, troubleshooting partner, or study tool can maintain a longer conversation without abruptly being asked to wait. For Windows users, that can mean keeping a multi-step PowerShell, Registry, deployment, or driver-debugging discussion in one thread rather than re-explaining system details after a cap resets.
It does not make all ChatGPT features unlimited. OpenAI explicitly says file uploads, image generation and analysis, and other tools remain subject to limits. The company has also attached an abuse-guardrail qualification to unlimited text chats, but has not published the thresholds, behaviors, regions, or enforcement conditions that could trigger a restriction. “Unlimited” therefore describes ordinary text use, not unconditional access to every ChatGPT resource.
OpenAI’s own support documentation also preserves an important boundary: Luna and Terra are not selectable in standard ChatGPT conversations. Free and Go users receive Luna as the default experience, while paid users have broader choices in ChatGPT Work, Codex, and the API. Users looking through the model picker for a Luna button may not find one even after the rollout reaches their account.
That leaves free users with a useful but bounded upgrade. Reasoning effort can improve a model’s handling of multi-step tasks, planning, calculations, and tricky instructions, but it does not turn a lower-cost model into the flagship tier. The underlying capability, tool behavior, and available quota still depend on the plan and product.
This is especially relevant to technical support use. A Think-enabled Luna response may do a better job tracing why a Windows update deployment failed, organizing a remediation plan, or checking a multi-part configuration change than a fast default response. It should not be treated as a substitute for validation against Event Viewer logs, Microsoft documentation, device-management telemetry, or a test machine. Higher reasoning reduces the chance of a shallow answer; it does not convert generated guidance into an authoritative record.
OpenAI is also staggering the rollout. Luna becomes the Free and Go default “this week,” while unlimited text chats and Think are scheduled for “next week.” Since the announcement is dated Thursday, August 6, OpenAI has not supplied a firm calendar date for the second stage. Accounts that see Luna now should not assume their text limit or Think button has already changed.
The rollout language is more consequential than it looks. OpenAI’s support page says availability is gradual, and managed workspaces can be governed by administrators. Free users should expect account-by-account variation during the deployment rather than a single global switch.
The stated goal is to reduce the sensation of changing personalities when a conversation moves from an instant answer to a deliberate one. OpenAI says the revised Sol powers both modes in the Chat experience, so the difference should be the amount of work performed rather than a change to another model with different style and behavior.
That is a better control surface for users who understand the trade-off. A lightweight question—such as the syntax of a
OpenAI says Sol’s revised behavior is more direct, uses less unnecessary formatting, and is more willing to correct a faulty assumption instead of simply agreeing. Those are behavior promises, not independently reproducible performance results. The company’s numerical reliability claim also deserves to be read carefully: in an internal set of financial, medical, and legal prompts, OpenAI says responses containing at least one factual error were 62% less common with Luna and 68% less common with Sol than with GPT‑5.5 Instant.
The announcement does not provide the test set, sample size, scoring protocol, source-access conditions, or absolute error rates. A 68% relative reduction could be major or modest depending on how often the previous model failed. It is evidence that OpenAI measured an improvement, but not enough information for administrators or regulated users to calculate operational risk from the figure.
That is broader than the product announcement, which repeatedly names Plus and Pro without explaining whether Business and Enterprise receive the new slider simultaneously. The Help Center adds another practical caveat: Enterprise customers may continue to see the existing model picker until the new controls reach their workspace. For managed Windows environments, the interface an employee sees may therefore lag behind the consumer-facing release.
There is a second documentation problem. In the same Help Center article, an “Instant limits” section still says Free-tier users have limited GPT‑5.5 Instant access in a five-hour window. Elsewhere on that page, OpenAI says Free and Go users are moving to GPT‑5.6 Luna, and the main announcement says unlimited Luna text chats begin next week.
Those statements may reflect a transition period or simply an unedited legacy section. They do not establish that OpenAI has quietly retained the old GPT‑5.5 limit for Luna. They do establish that the public documentation is currently ambiguous at the exact point users will care about most: when unlimited access begins and what guardrails can override it.
That means organizations should not infer a model refresh in their coding workflows, API integrations, internal automations, or managed Work environments from the consumer ChatGPT announcement. Developers retain separate Sol, Terra, and Luna API options, with their own prices, tool support, model snapshots, rate limits, and deployment decisions. A better everyday-chat response style does not automatically change an API workload’s behavior.
For IT teams, the immediate action is administrative rather than technical: check the actual ChatGPT model picker and workspace controls before revising support guidance or internal AI-use documentation. Free users should look for Luna first, then wait for the separate unlimited-text and Think rollout. Paid users should test the Sol slider with representative tasks before assuming that a more concise answer is also a more complete one.
OpenAI’s upgrade removes a major barrier for free text conversations, but it does so by scaling a low-cost model and keeping the flagship reasoning tier behind paid plans. The date that will matter is not August 6, when the announcement appeared, but the still-unspecified point next week when the Free and Go text allowance and Think button actually arrive.
OpenAI dated the announcement August 6, 2026. The supplied publication date of August 4 does not match the company’s live record. Axios independently reported the rollout on August 6 and confirmed the broad outlines: Luna becomes the default for Free and Go, while paid accounts receive the Sol-focused ChatGPT changes.
GPT‑5.6 Luna Is a Free-Tier Capacity Upgrade, Not a Flagship Upgrade
OpenAI introduced GPT‑5.6 as a three-tier family in July: Sol for its highest-capability work, Terra as a middle option, and Luna as the speed- and cost-oriented model. Its developer documentation describes Luna as designed for cost-sensitive, high-volume workloads—language that helps explain why it, rather than Sol, can underpin a much larger free text-chat allowance.That makes this announcement as much about serving cost as model quality. OpenAI cut Luna’s API price shortly after the GPT‑5.6 family launched; the current published rate is $0.20 per million input tokens and $1.20 per million output tokens. A model cheap enough to run in high volume gives OpenAI room to remove the most visible free-tier friction: the message cap in ordinary text conversations.
For casual use, that is a meaningful change. A person using ChatGPT as a drafting assistant, search companion, troubleshooting partner, or study tool can maintain a longer conversation without abruptly being asked to wait. For Windows users, that can mean keeping a multi-step PowerShell, Registry, deployment, or driver-debugging discussion in one thread rather than re-explaining system details after a cap resets.
It does not make all ChatGPT features unlimited. OpenAI explicitly says file uploads, image generation and analysis, and other tools remain subject to limits. The company has also attached an abuse-guardrail qualification to unlimited text chats, but has not published the thresholds, behaviors, regions, or enforcement conditions that could trigger a restriction. “Unlimited” therefore describes ordinary text use, not unconditional access to every ChatGPT resource.
OpenAI’s own support documentation also preserves an important boundary: Luna and Terra are not selectable in standard ChatGPT conversations. Free and Go users receive Luna as the default experience, while paid users have broader choices in ChatGPT Work, Codex, and the API. Users looking through the model picker for a Luna button may not find one even after the rollout reaches their account.
The Think Button Uses More Luna, Not Sol
The new Think control is designed to let Free and Go users ask Luna to use higher reasoning effort for a harder question. OpenAI’s Help Center is unambiguous on this point: Think uses GPT‑5.6 Luna, not GPT‑5.6 Sol.That leaves free users with a useful but bounded upgrade. Reasoning effort can improve a model’s handling of multi-step tasks, planning, calculations, and tricky instructions, but it does not turn a lower-cost model into the flagship tier. The underlying capability, tool behavior, and available quota still depend on the plan and product.
This is especially relevant to technical support use. A Think-enabled Luna response may do a better job tracing why a Windows update deployment failed, organizing a remediation plan, or checking a multi-part configuration change than a fast default response. It should not be treated as a substitute for validation against Event Viewer logs, Microsoft documentation, device-management telemetry, or a test machine. Higher reasoning reduces the chance of a shallow answer; it does not convert generated guidance into an authoritative record.
OpenAI is also staggering the rollout. Luna becomes the Free and Go default “this week,” while unlimited text chats and Think are scheduled for “next week.” Since the announcement is dated Thursday, August 6, OpenAI has not supplied a firm calendar date for the second stage. Accounts that see Luna now should not assume their text limit or Think button has already changed.
The rollout language is more consequential than it looks. OpenAI’s support page says availability is gradual, and managed workspaces can be governed by administrators. Free users should expect account-by-account variation during the deployment rather than a single global switch.
Sol’s Slider Replaces a Model-Switching Mental Model
For Plus and Pro subscribers, OpenAI’s more consequential interface change is the reasoning slider. It lets the user choose how much thought GPT‑5.6 Sol applies to a reply, from quick handling for ordinary questions to greater effort for planning, research, writing, code, and decisions.The stated goal is to reduce the sensation of changing personalities when a conversation moves from an instant answer to a deliberate one. OpenAI says the revised Sol powers both modes in the Chat experience, so the difference should be the amount of work performed rather than a change to another model with different style and behavior.
That is a better control surface for users who understand the trade-off. A lightweight question—such as the syntax of a
winget command or whether a Windows feature is supported—does not benefit much from a long internal analysis. A request to compare Intune remediation approaches, review a deployment script, map a Windows 11 migration plan, or isolate an intermittent crash does. The slider makes that choice visible instead of leaving the product to decide silently when a prompt deserves more inference time.OpenAI says Sol’s revised behavior is more direct, uses less unnecessary formatting, and is more willing to correct a faulty assumption instead of simply agreeing. Those are behavior promises, not independently reproducible performance results. The company’s numerical reliability claim also deserves to be read carefully: in an internal set of financial, medical, and legal prompts, OpenAI says responses containing at least one factual error were 62% less common with Luna and 68% less common with Sol than with GPT‑5.5 Instant.
The announcement does not provide the test set, sample size, scoring protocol, source-access conditions, or absolute error rates. A 68% relative reduction could be major or modest depending on how often the previous model failed. It is evidence that OpenAI measured an improvement, but not enough information for administrators or regulated users to calculate operational risk from the figure.
OpenAI’s Documentation Has Not Fully Caught Up
OpenAI’s updated GPT‑5.6 Help Center article confirms that Luna is becoming the Free and Go default and that Think will arrive next week. It also says the Sol slider is rolling out to eligible paid plans, while its plan table lists Medium and High Sol reasoning for Plus, Pro, Business, and Enterprise accounts.That is broader than the product announcement, which repeatedly names Plus and Pro without explaining whether Business and Enterprise receive the new slider simultaneously. The Help Center adds another practical caveat: Enterprise customers may continue to see the existing model picker until the new controls reach their workspace. For managed Windows environments, the interface an employee sees may therefore lag behind the consumer-facing release.
There is a second documentation problem. In the same Help Center article, an “Instant limits” section still says Free-tier users have limited GPT‑5.5 Instant access in a five-hour window. Elsewhere on that page, OpenAI says Free and Go users are moving to GPT‑5.6 Luna, and the main announcement says unlimited Luna text chats begin next week.
Those statements may reflect a transition period or simply an unedited legacy section. They do not establish that OpenAI has quietly retained the old GPT‑5.5 limit for Luna. They do establish that the public documentation is currently ambiguous at the exact point users will care about most: when unlimited access begins and what guardrails can override it.
Work, Codex, and API Users Are Not Being Moved With Chat
OpenAI has drawn a hard product boundary around this release. The revised GPT‑5.6 Sol is optimized for everyday ChatGPT conversations and is available only in the Chat experience. The Sol version used by ChatGPT Work and Codex is not changing as part of this rollout.That means organizations should not infer a model refresh in their coding workflows, API integrations, internal automations, or managed Work environments from the consumer ChatGPT announcement. Developers retain separate Sol, Terra, and Luna API options, with their own prices, tool support, model snapshots, rate limits, and deployment decisions. A better everyday-chat response style does not automatically change an API workload’s behavior.
For IT teams, the immediate action is administrative rather than technical: check the actual ChatGPT model picker and workspace controls before revising support guidance or internal AI-use documentation. Free users should look for Luna first, then wait for the separate unlimited-text and Think rollout. Paid users should test the Sol slider with representative tasks before assuming that a more concise answer is also a more complete one.
OpenAI’s upgrade removes a major barrier for free text conversations, but it does so by scaling a low-cost model and keeping the flagship reasoning tier behind paid plans. The date that will matter is not August 6, when the announcement appeared, but the still-unspecified point next week when the Free and Go text allowance and Think button actually arrive.
References
- Primary source: OpenAI
Published: 2026-08-04T00:00:00+00:00
Loading…
openai.com - Related coverage: axios.com
Loading…
www.axios.com - Related coverage: help.openai.com
Model Release Notes | OpenAI Help Center
help.openai.com
- Related coverage: developers.openai.com
GPT-5.6 Luna Model | OpenAI API
developers.openai.com
- Related coverage: openai.com
GPT-5.6: Frontier intelligence that scales with your ambition | OpenAI
More intelligence from every token, stronger performance per dollar, and more capability on demand for your hardest work.
openai.com
- Related coverage: help.openai.com
GPT-5.6 in ChatGPT | OpenAI Help Center
Learn how GPT-5.6 Sol works in ChatGPT, which options are available by plan, and how usage limits and availability work.
help.openai.com
- Related coverage: academy.openai.com
Loading…
academy.openai.com - Related coverage: deploymentsafety.openai.com
- Related coverage: axios.com
OpenAI discounts GPT-5.6 Luna and Terra, but not Luna
Efficiency gains drive major GPT-5.6 discounts.www.axios.com
- Related coverage: techradar.com
OpenAI’s next ChatGPT-5.6 upgrade may be too powerful to launch like a normal app update | TechRadar
There’s going to be a staggered rollout, starting Thursdaywww.techradar.com - Related coverage: cincodias.elpais.com
Loading…
cincodias.elpais.com