GPT-6 Sol Limits Across ChatGPT Plans and API Tiers

GPT-6 Sol limits depend on the ChatGPT plan, on whether the work runs in Chat or in Work and Codex, and on the application programming interface (API) usage tier. OpenAI deployed the model on September 22, 2026, as a lower-cost alternative to GPT-6 Astra, and expanded it to ChatGPT Chat on October 7, 2026. OpenAI publishes API rate limits, the context window and estimated Codex message ranges, but no fixed message limit for GPT-6 Sol in ChatGPT Chat.

GPT-6 Sol powers GPT-6 in the ChatGPT Chat tab for Plus, Pro, Business and Enterprise at Instant, Medium, High and Extra High, though Plus does not get Extra High. Free and Go tiers do not receive access to GPT-6 Sol. At a Thinking limit, ChatGPT shows when the allowance resets if that information is available, and it may continue with another available model.

In the API, OpenAI sets GPT-6 Sol rate limits at the organization and project level, under three paid usage tiers: Build, Launch and Grow. The model supports a context window of 1,050,000 tokens and a maximum output ceiling of 128,000 tokens per request.

GPT-6 Sol Limits by Plan

OpenAI sets varying GPT-6 Sol usage thresholds across ChatGPT subscription tiers, with published numeric estimates in Codex and unpublished message caps in standard Chat. OpenAI states that usage limits depend on the plan and the model. API requests instead have published rate limits, set by usage tiers that rise with an organization's total credit purchases.

In the primary ChatGPT Chat workspace, GPT-6 Sol powers the Instant, Medium, High, and Extra High effort settings for paid tiers. Free and Go accounts do not get GPT-6 Sol. OpenAI's October 7 post says they get GPT-6 Luna, while the OpenAI Help Center says GPT-5.6 Luna. Subscribers on ChatGPT Plus can access Medium and High reasoning, while Pro, Business, and Enterprise users gain access to Extra High reasoning.

ChatGPT shows when an allowance resets, when that information is available. OpenAI Support does not reset ChatGPT or Codex usage limits.

• Free ($0): no GPT-6 Sol. Chat runs on GPT-6 Luna per OpenAI's October 7 post, or GPT-5.6 Luna per the help center.
• Go ($8 per month): no GPT-6 Sol. Chat runs on GPT-6 Luna or GPT-5.6 Luna, the same split as Free.
• Plus ($20 per month): GPT-6 Sol at Instant, Medium and High, with Chat message limits not published. Work and Codex: an estimated 15 to 150 local messages per five hours.
• Pro 100 ($100 per month): GPT-6 Sol at Instant through Extra High, with limits not published. Pro effort runs GPT-6 Pro on GPT-6 Astra. Codex has no five-hour limit, and weekly limits may apply.
• Pro 200 ($200 per month): GPT-6 Sol at Instant through Extra High, with limits not published. Codex usage is higher than Pro 100 with no five-hour limit, though new subscriptions not eligible for grandfathering get a lower allowance. When the weekly GPT-6 Pro limit runs out, ChatGPT switches to GPT-5.6 Thinking at Medium.
• Pro 500 ($500 per month): GPT-6 Sol at Instant through Extra High, with limits not published. Codex has no five-hour limit, and Pro 500 is the only tier with Ultrafast, which runs on GPT-6.1 Sol and GPT-6 Astra but not GPT-6 Sol.
• Business Standard ($20 per user per month): Instant chat is virtually unlimited, but Medium, High and Extra High limits are not published. Work and Codex: an estimated 15 to 150 local messages per five hours. The seat also gets 15 Pro messages per month, which run on GPT-6 Pro, not GPT-6 Sol.
• Business Premium: Instant chat is virtually unlimited, and reasoning limits are not published. Codex gets five times the Standard seat's usage with no five-hour limit. The seat also gets 50 Pro messages per week on GPT-6 Pro, not GPT-6 Sol.
• Enterprise and Edu: Enterprise offers virtually unlimited messages for eligible base models, which OpenAI says does not mean zero-dollar usage, and admins can turn off GPT-6 Sol access. In Codex, flexible-pricing seats have no fixed rate limits and scale with credits, while other seats get Plus-level limits for most features.

Context Window and Output Token Limits

GPT-6 Sol provides a context window of 1,050,000 tokens and a maximum output ceiling of 128,000 tokens per request. The model accepts text and image inputs, while producing text-only outputs. GPT-6 Sol does not accept audio or video.

Requests that exceed 272,000 input tokens trigger OpenAI's long-context pricing rules. When a prompt crosses this threshold, OpenAI charges two times the standard input and cache token rates, along with 1.5 times the output rate for the full request. Standard GPT-6 Sol API pricing is $2.00 per million input tokens, $0.20 per million cached input tokens and $10.00 per million output tokens.

Supported API endpoints include Chat Completions (v1/chat/completions), Responses (v1/responses), and Batch (v1/batch). Unsupported interfaces include Realtime, Live, Assistants, Fine-tuning, Embeddings, image generation, and speech endpoints. Through the Responses API, GPT-6 Sol supports web search, file search, code interpreter, computer use, Model Context Protocol (MCP), and tool search.

• Maximum context window: 1,050,000 tokens per API call.
• Maximum output capacity: 128,000 tokens per generation cycle.
• Long-context surcharge threshold: Prompts above 272,000 input tokens incur 2x input and 1.5x output billing.
• Function calling: works in Chat Completions only with reasoning effort set to none. Tool use with reasoning needs the Responses API.

API Rate Limits and Usage Tiers

OpenAI governs gpt-6-sol API requests using organization-level rate tiers that scale automatically based on cumulative credit purchases. Rate limits apply across both organization and project scopes rather than individual user seats. Free tier API access is marked as not supported for gpt-6-sol.

The three paid usage tiers are Build, Launch, and Grow. The Build tier starts at $5 in total credit purchases and has a $500 monthly usage limit. It allows 5,000 requests per minute (RPM), 1,000,000 tokens per minute (TPM) and a 3,000,000 batch queue limit. The Launch tier requires $100 in purchases, raising limits to 10,000 RPM, 4,000,000 TPM, and a 200,000,000 batch queue limit, with a $5,000 monthly ceiling.

The Grow tier starts at $500 in credit purchases and has a $200,000 monthly usage limit, with 15,000 RPM, 40,000,000 TPM and a 15,000,000,000 batch queue limit. OpenAI's rate-limits guide gives Astra, Sol and Terra the same Standard limits, and GPT-6 Luna limits are higher. On price, Batch and Flex cost 50 percent of Standard rates and Fast mode costs 2x.

• Free Tier: gpt-6-sol access is not supported.
• Build Tier ($5 spent): 5,000 RPM, 1,000,000 TPM, and 3,000,000 batch queue limit.
• Launch Tier ($100 spent): 10,000 RPM, 4,000,000 TPM, and 200,000,000 batch queue limit.
• Grow Tier ($500 spent): 15,000 RPM, 40,000,000 TPM, and 15,000,000,000 batch queue limit.

Threshold Behaviors and Automated Fallbacks

ChatGPT and Codex handle a used-up GPT-6 Sol allowance differently. In ChatGPT Chat, hitting a Thinking limit at Medium, High or Extra High means ChatGPT shows when the allowance resets, if that information is available. ChatGPT may also continue the conversation with another available model.

In ChatGPT Work and Codex, an agent that reaches its usage limits during an active turn can keep working on that turn, subject to fair use limits. Plus and Pro users can buy credits.

On Business, automatic reasoning may use the reasoning allowance even while Instant is selected. The control is under Settings > General > Higher intelligence. On Pro $200 tiers, exceeding weekly GPT-6 Pro limits triggers an automatic fallback to GPT-5.6 Thinking at Medium reasoning effort.

ChatGPT Work and Codex Message Allowances

ChatGPT Work and Codex share usage, so Work inside ChatGPT draws on the same pricing, credits and usage limits as Codex, per OpenAI's Codex pricing page.

For Plus and Standard Business subscribers, OpenAI publishes estimated local message allowances across five-hour operating periods. GPT-6 Sol provides an estimated 15 to 150 local messages per five-hour window. By comparison, the Codex pricing page gives GPT-6.1 Sol 15 to 160 local messages, though OpenAI's Business help article groups it with GPT-6 Sol at 15 to 150. GPT-6 Astra gets 5 to 45, and GPT-6 Luna gets 350 to 3,000. OpenAI emphasizes that these numbers are operational estimates rather than fixed caps.

Pro plans have no five-hour limit in Codex, though weekly limits may also apply, and OpenAI publishes no numeric Pro allowance for GPT-6 Sol. Credit consumption rates for GPT-6 Sol at Standard execution speed are 50 credits per million input tokens, 5 credits per million cached input tokens, and 250 credits per million output tokens. Fast mode consumes included plan allowances at 2.5 times standard rates and purchased credits at double rates.

• GPT-6 Sol local estimate: 15 to 150 messages per five hours for Plus and Standard Business.
• GPT-6.1 Sol local estimate: 15 to 160 messages per five hours on the Codex pricing page, or 15 to 150 in OpenAI's Business help article.
• GPT-6 Astra local estimate: 5 to 45 messages per five hours.
• Standard credit consumption: 50 input, 5 cached input, and 250 output credits per million tokens.

Model Allocation Across Plan Tiers

Since OpenAI's October 7, 2026 ChatGPT update, different ChatGPT plans get different GPT-6 models. In the standard Chat tab, GPT-6 is powered by GPT-6 Sol for Plus, Pro, Business, and Enterprise plans. In the interface model menu, the entry is labeled simply as GPT-6, omitting the Sol designation.

Free and Go get GPT-6 Luna according to OpenAI's October 7 post, or GPT-5.6 Luna according to the help center article updated in October 2026. Thinking levels vary by subscription, with Plus users limited to Medium and High effort, while Pro, Business, and Enterprise accounts can select Extra High effort. Selecting Pro effort bypasses GPT-6 Sol entirely to run on GPT-6 Astra.

Technical capabilities also separate GPT-6 Sol from sibling models. GPT-6 Sol supports reasoning effort configurations of none, low, medium, high, xhigh, and max, whereas GPT-6.1 Sol and GPT-6 Astra do not support the none setting. OpenAI's model pages give GPT-6 Sol a knowledge cutoff of April 20, 2026, against April 30, 2026 for GPT-6.1 Sol and GPT-6 Astra and May 18, 2026 for GPT-6 Luna.

Operational Boundaries and Upgrade Conditions

GPT-6 Sol does not serve development teams requiring real-time audio streaming, video inputs, model fine-tuning, or native embeddings through the API. Those jobs need a different OpenAI model that supports the Realtime, fine-tuning or Embeddings endpoints. Teams that want Ultrafast mode should look at GPT-6.1 Sol or GPT-6 Astra, because OpenAI offers Ultrafast on those two and not on GPT-6 Sol.

This limits assessment would change if OpenAI publishes fixed numeric message ceilings for ChatGPT Chat, or introduces an official deprecation date for gpt-6-sol in the API documentation.

If you use the API, compare your organization's usage tier with the GPT-6 Sol limits in OpenAI's rate-limits guide before you move a high-volume job onto gpt-6-sol.