Codex Usage
Back to Codex Usage Overview

Official model release · Usage estimates verified 03 Oct 2026

GPT-6.1 Sol: Availability, Codex CLI Default & Usage Limits

GPT-6.1 Sol is real and is rolling out in ChatGPT Work and Codex. Released by OpenAI on September 29, 2026, gpt-6.1-sol delivers near-Astra reasoning performance for complex coding and agentic workflows at lower operating cost than Astra. Codex CLI 0.159.1 configured GPT-6.1 Sol as the default model in its bundled catalog.

GPT-6.1 Sol is available or rolling out to eligible paid plans, including Plus, Pro ($100, $200, and $500), Business (Standard and Premium), Enterprise, and Edu. Enterprise and Edu workspaces require an administrator to enable access. In ChatGPT, access is granted specifically inside Work and Codex surfaces—not standard personal Chat.

Selecting GPT-6.1 Sol in Codex

Codex CLI 0.159.1 sets gpt-6.1-sol as the default model in the bundled catalog. If your client is updated and your account has received the rollout, you can invoke or configure it through three methods:

1. Command-line flag

Pass the model directly when starting a session:

codex --model gpt-6.1-sol
2. In-session model picker

Type the model command inside any active interactive session:

/model
3. Persistent configuration (config.toml)

Set the default for all Codex CLI, IDE extension, and desktop sessions:

model = "gpt-6.1-sol"

For broader model selection context and preset guidance across Astra, Sol, and Luna, see our recommended model guide.

GPT-6.1 Sol vs. Original GPT-6 Sol

OpenAI documents GPT-6.1 Sol as the newer Sol model. Both models remain active, supported, and available for selection. Key differences include:

DimensionGPT-6.1 Sol (gpt-6.1-sol)Original GPT-6 Sol (gpt-6-sol)
CapabilitiesNear-Astra performance for complex agentic workflowsStandard Sol reasoning for difficult coding tasks
CLI Default StatusDefault model in Codex CLI 0.159.1 catalogSelectable via picker and configuration
Reasoning Effort SupportLow, Medium, High, XHigh, Max (no None or Minimal)Includes lower reasoning options
Work & Codex Credit Rates50 credits in / 2.5 credits cached / 250 credits out50 credits in / 5 credits cached / 250 credits out
API Standard Cached Input$0.1 / 1M tokens$0.2 / 1M tokens
Published 5-Hour Estimates15 to 160 (Plus), 15 to 160 (Business Standard); estimated local messages15 to 150 (Plus), 15 to 150 (Business Standard); estimated local messages
Lifecycle StatusActive (introduced Sep 29, 2026)Active (introduced Sep 22, 2026)

For historical background and detailed generation comparisons, visit our original GPT-6 Sol reference page.

Credit Rates vs. OpenAI API Pricing

Codex usage operates under two completely different accounting systems depending on how you sign in: ChatGPT Work/Codex credits or OpenAI Platform API token billing. Never calculate subscription usage using API dollar prices.

ChatGPT Work & Codex Credits

These token rates apply when GPT-6.1 Sol usage is billed against ChatGPT Work and Codex credits. Included subscription allowances are a separate layer and are used first where applicable. Token rates per 1,000,000 tokens at Standard speed:

  • • Input: 50 credits
  • • Cached input: 2.5 credits
  • • Cache writes: 0 credits (not charged)
  • • Output: 250 credits

OpenAI says a typical GPT-6.1 Sol Codex task may consume approximately 2 to 15 credits. Actual consumption varies with task complexity, input and cached-input size, output length, automations, reasoning behavior, and speed settings.

OpenAI Platform API (USD)

Applies when authenticating with an API key (OPENAI_API_KEY). Billed in USD per 1,000,000 tokens:

  • • Input: $2.00
  • • Cached input: $0.10
  • • Cache write: $2.50
  • • Output: $10.00

Inputs over 272K tokens incur 2x input and cache rates and 1.5x output rates. Fast mode is 2x Standard where supported; Batch and Flex tiers are 50% below Standard.

For more on billing layers, see Codex pricing and how credits work.

Subscription Limits: What Is and Isn't Published

OpenAI estimates 15 to 160 local GPT-6.1 Sol messages per five-hour window for Plus and 15 to 160 for Business Standard. These are estimates, not guaranteed message counts. Usage varies with task, model, and settings; cloud tasks may consume more allowance than local messages, and weekly limits may also apply. OpenAI usage guidance publishes these ranges.

Do not infer that GPT-6.1 Sol consumes your allowance at the exact same rate as GPT-6 Sol merely because their credit rates are similar. Furthermore, API token pricing is not a formula for your ChatGPT subscription allowance.

OpenAI does not publish a corresponding GPT-6.1 Sol Pro message estimate. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex. That does not mean unlimited included usage: these tiers still have included allowances. Check Settings > Usage for remaining allowance and reset times. Business Premium also has no five-hour cap, while some Enterprise and Edu workspaces use flexible credit-based pricing without fixed rate limits. Switching models does not reset the shared Work and Codex allowance; watch your overall Codex usage as you work.

To understand 5-hour rolling resets, weekly reset anchors, and tier allowances, read our canonical Codex limits reference.

Context Window and Max Output Tokens

OpenAI's API specification for gpt-6.1-sol defines a 1,050,000-token context window with up to 128,000 maximum output tokens.

API context specifications represent the maximum theoretical token capacity of the model, not a guaranteed client context size or a subscription allowance. Codex clients manage context window compaction, tool output trimming, and conversation history independently.

Read how Codex handles context windows for details on client compaction and token management.

Frequently Asked Questions

Is GPT-6.1 Sol real and officially released?

Yes. OpenAI released GPT-6.1 Sol on September 29, 2026. The official model identifier is gpt-6.1-sol. It is rolling out to paid plans in ChatGPT Work and Codex, and Codex CLI 0.159.1 bundled it as the default model in its local catalog.

How do I use GPT-6.1 Sol in the Codex CLI?

Update your Codex CLI to version 0.159.1 or newer. You can launch sessions with 'codex --model gpt-6.1-sol', switch interactively via the '/model' command, or set 'model = "gpt-6.1-sol"' in your config.toml file.

Does GPT-6.1 Sol replace or retire GPT-6 Sol?

No. While OpenAI documents GPT-6.1 Sol as the newer Sol model with near-Astra capabilities, original GPT-6 Sol (gpt-6-sol) remains active, supported, and selectable without a scheduled sunset date.

How many messages can I send with GPT-6.1 Sol on Plus or Pro?

OpenAI estimates 15 to 160 local GPT-6.1 Sol messages per five-hour window for Plus and 15 to 160 for Business Standard. These are estimates, not guaranteed counts; usage varies with task, model, and settings, cloud tasks may use more allowance, and weekly limits may also apply. OpenAI does not publish a corresponding GPT-6.1 Sol Pro message estimate. Pro $100, $200, and $500 currently have no five-hour usage limit in Work and Codex, but still have included usage allowances. Check Settings > Usage for remaining allowance and reset times. API usage is billed separately.

What reasoning effort levels does GPT-6.1 Sol support?

GPT-6.1 Sol supports low, medium, high, xhigh, and max reasoning effort. Unlike original GPT-6 Sol, GPT-6.1 Sol does not support 'none' or 'minimal' reasoning levels.

How do credits work for GPT-6.1 Sol?

In ChatGPT Work and Codex, GPT-6.1 Sol meters at 50 input credits, 2.5 cached input credits, and 250 output credits per million tokens. Cache writes are not charged on the credit rate card. OpenAI indicates typical tasks consume 2 to 15 credits, though actual consumption varies by prompt and codebase context.

Official sources: OpenAI GPT-6.1 Sol API Model Documentation, OpenAI Model Selection & Guidance, OpenAI Work & Codex Pricing.

Page last updated and verified against official sources: 2026-10-03T00:00:00Z.