Where it stands
OpenAI says GPT-6 Sol and Luna are available in the API and selected ChatGPT Work and Codex plans, with API prices 50% below GPT-5.6 promotional rates. The performance gains are OpenAI's own evaluations. Earlier leaks and posts, including a 750-tokens-per-second figure, did not confirm a launch, and the official material does not mention Terra or that speed. Later OpenAI material adds GPT-6.1 Sol, an Ultrafast tier at six times Standard pricing, and EU residency for Fast modes. Watch the gradual ChatGPT rollout and independent tests of the reported gains.
OpenAI adds an Ultrafast speed tier for GPT-6.1 Sol at six times Standard pricing
Ultrafast is OpenAI's new speed tier for GPT-6.1 Sol. OpenAI is rolling out Ultrafast for GPT-6.1 Sol in the API, Codex and ChatGPT Work, the company said on X on 8 October 2026. The company describes it as near-Astra intelligence at up to 8x the speed of Sol Standard. The API price is $12 per million input tokens and $60 per million output tokens, six times Sol Standard's $2 and $10 rates.
Ultrafast API pricing
OpenAI's longer announcement on its developer community forum gives the Ultrafast price as $12 per million input tokens and $60 per million output tokens. The company says this is 1.2 times the cost of Astra.
The Standard rates in this article are $2 per million input tokens and $10 per million output tokens. Against those rates, Ultrafast costs six times as much for both input and output. As a hypothetical example, a workload of 1 million input tokens and 1 million output tokens would cost $12 at Standard rates ($2 + $10) and $72 on Ultrafast ($12 + $60).
Fast mode, which costs twice the Standard price, is a separate option. Ultrafast has its own rate and should not be read as a version of Fast mode.
Who can use it
In Codex and ChatGPT Work, Ultrafast is listed for Pro 500, eligible usage-based Enterprise and credit-based Edu plans. Enterprise administrators must enable access for their organisation. The announcement names those plans, and this article reports only those.
In the API, the company says Ultrafast for GPT-6.1 Sol is available in all supported regions, including US and EU data residency.
EU data residency for Sol Fast
This article previously stated that Fast mode is unavailable with EU data residency. OpenAI's announcement says EU data residency support has now been added for GPT-6.1 Sol Fast and GPT-6 Luna Fast. Developers who avoided Fast mode for that reason should check availability in their own account.
Does this change your choice?
The decision depends on whether faster responses are worth roughly six times the Standard token price. OpenAI positions Ultrafast for work where speed matters, such as debugging an outage, agents that navigate apps and live experiences where each second counts. The company's "up to 8x" figure is a ceiling rather than a typical result.
In our view, batch jobs or workloads where latency matters little have less reason to pay the premium. The announcement does not include independent speed or quality benchmarks, so anyone deciding whether to use Ultrafast should measure latency and cost on their own workload.
Sources
OpenAI's developer docs detail GPT-6.1 Sol's context, tools and pricing
OpenAI Developers describes GPT-6.1 Sol as a model for complex coding, computer use and other professional work. Its documentation lists a 1,050,000-token context window, configurable reasoning, text and image input, and a range of API features.
The model page also includes important conditions around tool calling, long prompts, data residency, pricing modes and account tiers. Its performance and cost positioning should be evaluated on representative workloads rather than assumed from the documentation alone.
What GPT-6.1 Sol is designed to do
GPT-6.1 Sol is positioned for workloads that need more than a short text response, particularly complex coding, computer use and professional tasks. OpenAI describes it as offering “near-Astra performance at a lower cost,” but the model page does not provide independent benchmark scores or task-level results. That description is therefore OpenAI’s stated positioning, not a verified performance conclusion.
For an evaluation, the relevant question is whether the model’s reasoning options, long context, tool support and price fit the workload. A developer building a tool-using coding workflow, for example, could select the model ID gpt-6.1-sol and use the Responses API to combine model output with supported tools.
Reasoning, inputs and outputs
The reasoning.effort setting supports five levels:
lowmedium, the defaulthighxhighmax
The none and minimal settings are not supported. The available levels give developers a documented way to adjust reasoning effort for a task, although the model page does not quantify how quality, latency or token use changes at each setting.
GPT-6.1 Sol accepts text and image input and produces text output. Audio and video are not supported modalities. Image input could therefore be relevant to workflows that need the model to inspect visual material, while applications requiring native audio or video input would need a different model or an additional processing step.
The model has a context window of 1,050,000 tokens. The context window is the amount of input and related conversation or tool information the model can handle in a request. Its maximum output is 128,000 tokens. These are capacity limits, not a promise that every request will be processed at those sizes or that a longer response will be useful.
The documentation lists April 30, 2026 as the model’s knowledge cutoff.
API access and tool-use implications
The key distinction for tool-based applications is that tool calling is documented for the Responses API. Chat Completions is supported without tool calling.
The model page lists streaming, function calling and structured outputs as supported features. Function calling lets an application provide callable functions that the model can request, while structured outputs are intended for responses that follow a defined format.
Endpoint index and model capabilities
The documentation includes GPT-6.1 Sol in an endpoint index covering services such as:
Responses and Chat Completions
Live, Realtime, realtime translation and realtime transcription
Assistants and Batch
Embeddings, image generation, video and image-edit services
Speech generation, transcription and translation
Moderation and legacy Completions
This endpoint list should not be read as a list of every capability supported by GPT-6.1 Sol. The model-specific fields say that text is the only output modality, image is supported for input, and audio and video are not supported. They also say fine-tuning is not supported. The page does not explain the setup, quotas or model-specific behavior for each listed endpoint, so developers should confirm those details before designing around one.
Tools supported through the Responses API
The model page lists these tools as supported when GPT-6.1 Sol is used with the Responses API:
Web search
File search
Image generation
Code interpreter
Hosted shell
Apply patch
Skills
Computer use
MCP
Tool search
This combination is relevant to applications that need the model to work with files, execute code, interact with a computer environment or call external capabilities. The documentation does not provide task-level accuracy results for these tools, so support should not be confused with demonstrated reliability for a particular workflow.
GPT-6.1 Sol API pricing
The standard text-token rates listed in the documentation are:
Pricing item | Rate or adjustment | Qualification |
|---|---|---|
Input tokens | $2.00 per 1 million | Standard rate |
Cached input | $0.10 per 1 million | 5% of the uncached input rate |
Cache writes | $2.50 per 1 million | 1.25 times the uncached input rate |
Output tokens | $10.00 per 1 million | Standard rate |
Prompts above 272,000 input tokens | 2 times input and cache rates; 1.5 times output rate | Applies to the full request |
Fast mode | 2 times Standard | Unavailable with EU data residency |
Batch and Flex | 50% lower than Standard | Documented pricing adjustment |
Regional processing | 10% premium where available | Availability is conditional |
Cached input is charged at 5% of the uncached input rate, while cache writes cost 1.25 times the uncached input rate. These rates matter for applications that repeatedly send stable prompt material or manage long shared contexts.
When a prompt contains more than 272,000 input tokens, the documentation says the higher input and cache rates apply at twice the standard rate and the output rate rises to 1.5 times standard for the full request. Crossing the threshold can therefore affect the entire request rather than only the tokens beyond 272,000.
Fast mode costs twice the Standard price. Batch and Flex are listed at 50% below Standard, while regional processing adds a 10% premium where available. Fast mode is unavailable with EU data residency. Tool-specific models can also involve fees per tool call, according to the pricing note on the model page, so token totals alone may not describe the complete cost of a tool-using application.
Data residency and deployment constraints
GPT-6.1 Sol supports US and EU data residency. The documentation specifically notes that Fast mode cannot be used with EU data residency. Teams choosing a processing region should therefore consider both their data-handling requirements and whether the associated speed or pricing mode is available for that region.
The model page does not describe every residency eligibility condition or give endpoint-specific deployment instructions. Those details should be checked before committing a production workload to a particular processing arrangement.
Rate limits by usage tier
GPT-6.1 Sol is not supported on the Free tier. The listed limits for supported tiers are measured in requests per minute (RPM) and tokens per minute (TPM):
Usage tier | RPM | TPM |
|---|---|---|
Free | Not supported | Not supported |
Tier 1 | 500 | 500,000 |
Tier 2 | 5,000 | 1,000,000 |
Tier 3 | 5,000 | 2,000,000 |
Tier 4 | 10,000 | 4,000,000 |
Tier 5 | 15,000 | 40,000,000 |
The documentation says that usage tiers determine how high the limits are and can increase as an account sends more requests and spends more on the API. RPM limits affect request frequency, while TPM limits affect the total token volume an application can send or process over time.
These limits are important for capacity planning. A workload with large prompts may reach its token limit before its request-per-minute limit, while a high-volume application sending short requests may encounter the opposite constraint.
When might GPT-6.1 Sol be suitable?
GPT-6.1 Sol may be worth evaluating when an application needs several of the following:
Complex coding or professional-work capabilities
Reasoning effort above the default medium setting
Image input with text output
A context window of up to 1,050,000 tokens
Responses API tool calling
Computer use, code execution, web search or file search
Structured outputs or function calling
US or EU data residency
Batch or Flex pricing for workloads that can use those modes
The model may be a less obvious fit when an application requires native audio or video input, fine-tuning, Chat Completions tool calling, Fast mode with EU residency or Free-tier access. Its standard output price is also five times its uncached input price per million tokens, so applications that generate large responses should model output costs separately from prompt costs.
GPT-6.1 Sol evaluation checklist
Before selecting the model for production, developers should check:
Workload quality: Compare it with alternatives on representative coding, computer-use or professional tasks.
API design: Use the Responses API if the application depends on documented tool calling; Chat Completions is listed without tool calling.
Modalities: Confirm that text and image input with text output is sufficient, because audio and video are not supported modalities.
Context economics: Test whether prompts exceed 272,000 input tokens, since the higher pricing applies to the full request above that threshold.
Tool costs: Include any applicable per-tool-call charges alongside input and output token costs.
Capacity: Compare expected request and token volume with the account’s RPM and TPM tier limits.
Deployment requirements: Confirm US or EU residency needs and whether the selected pricing or speed mode is available in that region.
GPT-6.1 Sol’s documented combination of long context, configurable reasoning and Responses API tools makes it a candidate for demanding API workflows. The right choice depends on measured workload results, token mix, tool usage, latency requirements and deployment constraints.
Sources
OpenAI launches GPT-6 Sol and Luna in the API and selected ChatGPT plans
OpenAI has introduced GPT-6 Sol and GPT-6 Luna as faster, less expensive models for work that does not require GPT-6 Astra’s full capability. In a post on X, @OpenAI says the models bring advances from Astra to lower-cost use cases, with both available in the API and in selected ChatGPT Work and Codex plans.
The announcement also describes 50% lower API pricing than GPT-5.6 promotional pricing, improved prompt caching, and reported gains in automation, coding, factuality, computer use, communication, and alignment. These performance comparisons come from OpenAI’s evaluations or public reports about competitor models, not independent testing.
What are GPT-6 Sol and GPT-6 Luna?
GPT-6 Sol and GPT-6 Luna are positioned as the faster, more affordable members of OpenAI’s GPT-6 family. OpenAI says they were trained with methods similar to GPT-6 Astra and carry over advances in professional work, factuality, coding, computer use, and alignment.
Astra remains OpenAI’s highest-capability option in the announcement. Sol and Luna are instead aimed at situations where teams need more room to iterate, lower costs, or sustained use at scale. The announcement does not provide a complete technical specification comparison between Sol and Luna, so it does not establish a detailed division of workloads or capability limits between them.
The API model names are gpt-6-sol and gpt-6-luna.
GPT-6 Sol and Luna API pricing
OpenAI lists the following prices per 1 million tokens, compared with the promotional prices for the corresponding GPT-5.6 models:
Model | Comparison model | Input price: GPT-5.6 promotional → GPT-6 | Output price: GPT-5.6 promotional → GPT-6 | OpenAI’s stated reduction |
|---|---|---|---|---|
GPT-6 Sol | GPT-5.6 Sol | $4 → $2 | $20 → $10 | 50% cheaper |
GPT-6 Luna | GPT-5.6 Luna | $0.20 → $0.10 | $1.20 → $0.50 | 50% cheaper overall; the listed output reduction is approximately 58% |
OpenAI describes both models as 50% cheaper overall than their corresponding GPT-5.6 promotional prices. The listed Luna input price falls by 50%, while its output price falls from $1.20 to $0.50, or approximately 58%.
The lower prices are intended to make repeated experimentation and larger-scale workloads more practical. That matters especially for coding agents and business automation, where a task may involve many model calls rather than one isolated response.
The listed prices are API token prices. They should not be interpreted as a claim that every ChatGPT plan has unlimited access to either model.
Reported performance for automation, coding, and computer use
OpenAI groups Sol and Luna’s reported gains around tasks that require agents to work through multiple steps, use tools, or operate on real codebases.
On AutomationBench, which OpenAI describes as a test of business workflows across apps, GPT-6 Sol at xhigh effort scored 33.2% at a reported cost of $0.27 per task. OpenAI compares that result with 26.9% for Claude Opus 5 at maximum effort, while saying Sol cost 9% as much per task. GPT-6 Astra at low effort scored 30.3% in the same comparison, while Claude Fable 5.1 with an Opus 5 fallback scored 31.4% at more than 8.9 times Sol’s cost per task; the fallback cost was not reported.
At high effort, GPT-6 Luna improved on its predecessor by 5.4 percentage points at a reported 58% lower cost per task. OpenAI also says Sol exceeded Claude Fable 5.1 at a lower cost and surpassed low-effort GPT-6 Astra in the AutomationBench comparison.
On Agents’ Last Exam, which evaluates agents on complex professional workflows, OpenAI reports that GPT-6 Sol at maximum effort scored 56.4%. The company also reports that Sol cost 60% less per task than Claude Opus 5’s highest-scoring configuration.
Coding performance
OpenAI says GPT-6 Sol improves substantially over GPT-5.6 Sol on FrontierCode, an evaluation of whether coding agents produce changes ready to merge into real codebases. It also says Sol matched Claude Fable 5.1 at xhigh effort at a much lower cost. The announcement reports this comparison without providing the underlying FrontierCode scores.
On DeepSWE v1.1, which tests complex software-engineering work in real codebases, GPT-6 Sol at maximum effort scored 68.8%. OpenAI compares that with 69.9% for Claude Fable 5 at xhigh effort and reports that Sol cost approximately 80% less per task.
GPT-6 Luna at maximum effort scored 66.6% on the same evaluation. OpenAI describes that as comparable to Claude Opus 5 and Claude Fable 5 at medium effort, with Luna costing 93% less per task than Opus 5 and 96% less than Fable 5 in those comparisons.
Computer use
GPT-6 Astra remains the model OpenAI identifies as its strongest option for computer use. Even so, the company reports improved cost efficiency from the newer lower-cost models.
On OSWorld 2.0 offline, GPT-6 Sol at xhigh effort scored 60.5%, compared with 60.3% for Claude Opus 5 at medium effort. OpenAI says Sol achieved that similar result at approximately 80% lower cost per task. GPT-6 Luna at maximum effort also exceeded GPT-5.6 Sol at medium effort at one-tenth of the cost, according to the announcement.
These comparisons indicate how OpenAI frames the cost–performance trade-off. They do not establish that Sol or Luna will produce the same results on every production workload. OpenAI says its GPT evaluations were run in a research environment or through its API, which can differ from production ChatGPT because of system prompts, available tools, and other factors. It says competitor scores came from publicly available reports, and that Claude Fable 5 scores were used where Fable 5.1 scores were unavailable.
Factuality, communication, and alignment
OpenAI says GPT-6 Sol makes about half as many mistakes as its predecessor on an internal factuality evaluation based on de-identified real-world conversations in which users had flagged model mistakes. The company says Sol approaches Astra-level reliability at a much lower cost.
OpenAI also reports a substantial improvement for GPT-6 Luna. At higher effort levels, it says Luna matches GPT-5.6 Sol on that evaluation at about one-hundredth of the cost. The source does not include the evaluation’s sample size, full methodology, or independently reproduced results.
The models are also meant to produce clearer technical and coding responses. OpenAI describes the changes as less jargon, fewer unusual turns of phrase, fewer low-value details, and somewhat shorter answers without losing substance.
On alignment evaluations, OpenAI says both models improved over their GPT-5.6 counterparts, including lower rates of misleading statements about their coding work. The source describes these evaluations as deliberately challenging and says they do not measure failure rates in typical use. One supplied OpenAI chart labels this measure “coding deception” and reports lower rates for GPT-6 Sol and GPT-6 Luna than for their GPT-5.6 counterparts, but the announcement’s broader warning still applies: these are evaluation results, not guarantees of production behavior.
Prompt caching and lower costs for repeated context
The API changes are not limited to the headline token prices. OpenAI says GPT-6 models have higher prompt-cache hit rates by default and offer discounts of 90% on cached input-token reads.
Prompt caching lets an application reuse a previously processed portion of a prompt instead of paying the full processing price for the same context again. This can matter for agents that repeatedly send a system prompt, project instructions, tool definitions, or a large codebase context while changing only the latest user request.
OpenAI says developers can:
monitor cached input and changes over time in the Prompt Caching Dashboard;
use a diagnostics tool to identify missed caching opportunities;
change reasoning effort or enable and disable tools without breaking reuse of earlier context; and
set explicit cache breakpoints to control which prompt prefixes are cached.
The company also says GitHub observed that these improvements reduced the share of prompt tokens requiring fresh processing by more than 50% across billions of requests to OpenAI models over several months. That is a reported result about the cited usage, not a guaranteed reduction for every application.
Where GPT-6 Sol and Luna are available
OpenAI says both models are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app.
The announcement specifically says the models are not yet available in Chat. OpenAI also says the ChatGPT rollout is gradual, so users who do not immediately see the models in ChatGPT Work or Codex may need to try again later.
Both models are available through the OpenAI API. For API users, Sol and Luna offer lower listed token prices than the corresponding GPT-5.6 promotional prices, while the choice between them depends on the workload, required effort level, and the cost of repeated context. The announcement does not provide enough technical detail to define every capability difference between the two models, so developers should evaluate them against their own tasks rather than treating the model names alone as a workload recommendation.
For the source announcement and its full set of reported evaluations, see OpenAI’s introduction to GPT-6 Sol and Luna.
Sources
A post speculates on an Ultrafast mode and 750 tokens per second for GPT-6-Sol
A post on X by @kimmonismus speculates that a possible GPT-6-Sol release could include an “Ultrafast” mode and reach 750 tokens per second, where tokens are small text units an AI model processes or generates. The post does not confirm that GPT-6-Sol exists as a released model, that the mode is planned or available, or that the speed figure comes from a benchmark.
The post says: “No way, looks like we get Ultrafast mode on top of GPT-6-Sol release. That would be insane. GPT-6-Sol on 750t/s.” Its wording presents both details as speculation rather than as an announcement of a confirmed product.
What the post suggests
The message points to two separate ideas:
An Ultrafast mode: The post suggests that this could be offered alongside a GPT-6-Sol release. It provides no explanation of what the mode would do or how it would work, and does not say whether it would be available to users.
A speed of 750 tokens per second: The post associates this figure with GPT-6-Sol, but does not define how the number was measured.
What remains unconfirmed
There is no official announcement, release date, access information, pricing, model documentation or independent test result accompanying the post. It also does not establish whether GPT-6-Sol is a released model or whether “Ultrafast” is an actual mode rather than a prediction or reference to an unverified report.
What 750 tokens per second could mean
The post gives no measurement definition or testing context. It does not say whether 750 tokens per second refers to output generation, input processing, a short test, a particular hardware setup or another measurement. It also provides no benchmark method or comparison point.
That missing context matters because tokens are not necessarily the same as words, and a tokens-per-second figure is not meaningful on its own. A useful performance report would identify what was measured, the hardware or service conditions, the model configuration, the length of the test and how the result was calculated. The post itself provides no additional explanation of what the mode would do or how the speed figure was measured.
What would confirm the report
Confirmation would require evidence beyond the post, such as an official release notice or documentation identifying GPT-6-Sol and describing an Ultrafast mode. The 750-token-per-second figure would also need a reproducible benchmark that explains the test conditions and measurement method.
Until that evidence appears, the post is best understood as speculation about a possible GPT-6-Sol release, a possible Ultrafast mode and an unverified speed figure—not as confirmation that any of them exists.
Sources
A leaked Azure configuration lists GPT-6 Sol, Luna and Astra Minor names
An unverified post on X points to an attached screenshot that appears to show GPT-6 Sol, GPT-6 Luna and GPT-6 Astra Minor in a Microsoft Azure configuration excerpt. The image supports only the narrower observation that those names appear in registry-style paths; it does not establish that the models have launched, are functional or are available to subscribers.
The post also discusses GPT-5.6 Sol and mentions a possible Opus 5.5 release. The supplied evidence contains no Microsoft or OpenAI confirmation of a GPT-6 release or launch date.
What the X post reports
In a post on X, @bridgemindai says there could be a “double drop”: an Opus 5.5 release alongside the reported appearance of GPT-6 Sol, GPT-6 Luna and GPT-6 Astra Minor in Microsoft’s Azure configuration.
The post interprets the reported change as evidence that a replacement for GPT-5.6 Sol may be ready. It also suggests that the models could eventually be usable through existing subscriptions. Those are predictions in the post, not confirmed release details.
What the attached Azure configuration appears to show
The screenshot appears to show entries formatted as Azure ML registry paths, including:
azureml://registries/azure-openai/models/gpt-6-solazureml://registries/azure-openai/models/gpt-6-lunaazureml://registries/azure-openai/models/gpt-6-astra-minor
The visible excerpt places the three names alongside other model-style entries, including GPT-5.5 and GPT-5.6 Sol. That supports describing the image as a configuration or registry excerpt containing those names.
The image does not show what the entries do. It does not establish whether they refer to public models, internal references, experiments, placeholders or active deployments. The screenshot’s authenticity, provenance and creation time cannot be established from the supplied material, and it is unknown whether the listed entries are current or functional.
Why a registry entry is not the same as public availability
A name in a configuration can indicate that software is prepared to recognize a model or resource. By itself, it does not demonstrate that customers can select the model, send requests to it or access it through a subscription.
The screenshot provides no capabilities, pricing, access rules, geographic availability or release date for the three reported GPT-6 names. It also does not show that the models are ready for general use. Those details would require an official announcement or additional supporting documentation.
What the post says about GPT-5.6 Sol
The post says GPT-5.6 Sol was demoted to “workhorse” in the same commit. It does not define that label or explain how it would affect users. The screenshot’s apparent gpt-5.6-sol path likewise does not show that the model was replaced, withdrawn or moved to a particular tier.
The reported “workhorse” description therefore cannot establish that any GPT-6 model is available as a replacement. It is an interpretation of the change, not evidence of a confirmed product transition.
The separate Opus 5.5 mention
The post also mentions Opus 5.5, describing it as a possible release on the same day. The attached screenshot shows Azure-style entries with GPT names and provides no supporting details about Opus 5.5.
The Opus 5.5 reference therefore does not establish a connection with the reported GPT-6 entries. The post’s timing and release prediction remain unconfirmed.
Has Microsoft or OpenAI confirmed a release?
The supplied evidence contains no Microsoft or OpenAI confirmation of GPT-6 Sol, GPT-6 Luna, GPT-6 Astra Minor or a release date. The screenshot does not turn the names into an announcement, and it does not confirm public availability, capabilities or subscription access.
The supplied X reading is truncated, and the article linked from the post could not be read. For now, the image supports only the observation that three GPT-6-style names appear in what looks like a Microsoft Azure registry configuration excerpt. Whether those names represent planned products, internal testing, placeholders or something else remains unresolved.
Sources
A platform report points to preparation for GPT-6 Sol API and Codex access
In a post on X, @LuminaBench says it has seen preparation for GPT-6 Sol to become public, including entries for OpenAI API and Codex OAuth support. The post also points to image input, adjustable reasoning effort and long-context handling. Those observations suggest platform preparation, but they do not confirm that GPT-6 Sol has launched or that people can access it.
The report does not include an official OpenAI announcement, release date, public access instructions, API documentation or performance results. It also says pricing was not visible. The available evidence is therefore best understood as a brief report about apparent configuration work, not confirmation of a released model.
What the post reports
The post says it found multiple platforms adding what it describes as full GPT-6 Sol support. Its accompanying screenshot appears to show GPT-6 Sol in supported-model and display-name entries, alongside provider and runtime configuration fields.
The screenshot is evidence of what the post displays, not independent confirmation that the model or the listed routes work. Configuration entries can show preparation or compatibility intended by a platform without establishing that a model is publicly available.
Reported platform and provider paths
The reported entries include an openai-api-key path as the primary provider and an openrouter-codex path associated with openai/gpt-6-sol as a fallback. The screenshot also lists codex-oauth-token among the provider catalogs and refers to Codex fast-mode models.
An API key is a credential used by software to access an API. OAuth is an authorization method that lets an application obtain permission through an account sign-in flow, rather than requiring the application to handle a raw password. These labels indicate the kinds of integration the screenshot appears to describe; they do not establish that GPT-6 Sol access is available through either route.
Reported capabilities and configuration
The post mentions three model-related areas:
Image input: The screenshot appears to list GPT-6 Sol among models marked as supporting image input. That generally refers to sending an image alongside text for the model to process, but the post does not show a working request or describe supported image formats.
Reasoning effort: The displayed configuration refers to a range from low to ultra, with a maximum default. A reasoning-effort setting typically lets a platform trade response speed or resource use against the amount of internal processing requested, but no measured effect is supplied here.
Long context: The screenshot includes a long-context input-token setting with a displayed value of
272,001. Context is the text and other input information a model can consider during a request. The displayed number is a configuration detail in the screenshot, not independently verified documentation of a public context limit.
The screenshot also shows a $$$ built-in model price tier. The post says it still could not see pricing, and the screenshot describes the tier as display-only. That should not be treated as a confirmed price, rate card or estimate of what users would pay.
What remains unconfirmed
The supplied report does not establish any of the following:
an official OpenAI announcement or release date;
public access to GPT-6 Sol;
working OpenAI API, Codex OAuth or OpenRouter access;
API documentation, rate limits or access requirements;
confirmed pricing;
benchmark results, reliability data or comparative performance; or
the final capabilities, context limit or supported inputs of the model.
For now, the most precise reading is that a platform-related post reports apparent preparation for GPT-6 Sol across several integration and configuration areas. Readers should wait for primary documentation or an official announcement before treating those entries as proof of a launch or using them as access instructions.
Sources
A screenshot shows a gpt-6-sol label attributed to OpenAI
A reported screenshot has drawn attention to a model label called “gpt-6-sol.” In a post dated September 10, 2026, @mrfanduu shared the sighting, while the attached image displayed “gpt-6-sol” and a visible “Platform: OpenAI” field.
That is enough to document what the screenshot appears to show, but not enough to establish that GPT-6-sol is an officially announced or publicly available OpenAI model.
What the GPT-6-sol screenshot shows
The image contains three relevant pieces of visible text:
A truncated label beginning with “[NEW] New Model...”
The model name “gpt-6-sol”
“Platform: OpenAI”
The screenshot therefore presents GPT-6-sol as an entry associated with OpenAI within whatever interface produced the listing. However, the supplied image does not identify that interface, explain who created the entry, or provide additional information about the model.
Does the listing confirm an OpenAI model?
It identifies OpenAI as the platform shown in the screenshot, so the reported listing associates GPT-6-sol with OpenAI. That is different from confirmation that OpenAI has officially announced the model.
The post and image do not include an OpenAI statement, release information, documentation, access instructions, pricing, benchmarks, capabilities, or a launch date. They also do not establish whether the listing is authentic, experimental, internal, or the result of another source’s labeling.
What is GPT-6-sol?
Based on the available evidence, GPT-6-sol is a reported model label connected to an OpenAI platform listing. Its name and the visible platform attribution are the only substantive details currently supported by the supplied source.
Any interpretation of “sol,” its potential capabilities, or its relationship to other GPT models would go beyond what the screenshot confirms. For now, the item is best treated as a reported sighting rather than evidence of a released model.
A post predicts GPT-6 Sol, Terra and Luna model names
A short post from @Conor_D_Dart speculates that future models could be named GPT-6 Sol, GPT-6 Terra and GPT-6 Luna, with GPT-5.6 Luna also mentioned as a model to watch. The post presents these names as an expectation, not as a detailed product announcement.
What the post says
The prediction contains two related ideas: a future GPT-6 series using the names Sol, Terra and Luna, and a possible GPT-5.6 Luna release. It does not explain what the names represent, when any of the models might appear, or what capabilities they would have.
That makes the naming itself the only concrete subject of the post. Readers should treat the model names as speculation rather than confirmed roadmap details. Nothing in the post establishes a release date, technical specification or official availability.
0 comments
No approved comments yet. You can start the conversation.
Leave a comment