ChatGPT vs Grok (2026): $20 Plus vs $30 SuperGrok, Message Caps, Grok 4.7 vs GPT-6, and the Cheaper API

The Short Answer

For a consumer chat plan, ChatGPT is the cheaper entry point at every rung: $0 Free, $8 Go, $20 Plus. Grok's ladder is $10 SuperGrok Lite, $30 SuperGrok, $100 SuperGrok Plus, and $300 SuperGrok Heavy. ChatGPT Free is also more generous than it was in June. It moved to GPT-5.6 Luna with unlimited text chats in August 2026.

Grok's edge is live X data and API price. Grok 4.7 costs $2 input / $6 output per 1M tokens. GPT-6 Sol, released the next day, costs $2 / $10. The older gap was wider: Grok 4.3 at $2.50 output against GPT-5.5 at $30. Both labs cut prices, so "10x cheaper" is now roughly 1.7x on output between the two new flagships.

What changed since August 2026

• Sept 22: GPT-6 Sol and GPT-6 Luna ship in the API, ChatGPT Work, and Codex.
• Sept 21: Grok 4.7 ships on the xAI API, in Cursor, and in Grok Build. Grok 4.7 Fast is Cursor and Grok Build only.
• Sept 17: ChatGPT for Word launches on all plans, including Free. xAI ships Grok Voice Transcribe 2.0.
• Sept 14: OpenAI announces GPT-5.5 leaves ChatGPT, Work, and Codex on October 14, 2026.
• Sept 10: OpenAI pauses new sign-ups and upgrades to ChatGPT Pro $200.
• Sept 9: ChatGPT Voice daily limits: Go 3 hours on GPT-Live-1 mini, Plus 3 hours, Pro $100 15 hours, Pro $200 unlimited.
• Sept 3: GPT-6 Astra launches. In ChatGPT it appears as GPT-6 Pro with weekly caps.
• Aug 12: Grok 4.6 ships at $2 / $6 per 1M with a 500K context window.
• Aug 11: xAI launches Grok Bot, agents that run on a persistent cloud computer.
• Aug 6: ChatGPT Free and Go move to GPT-5.6 Luna with unlimited text chats.

Plan Pricing: ChatGPT Plus vs SuperGrok, Go vs SuperGrok Lite

ChatGPT undercuts Grok by $2 at the entry tier ($8 Go vs $10 SuperGrok Lite) and by $10 at the main tier ($20 Plus vs $30 SuperGrok). Grok prices in this table come from third-party trackers, because x.ai's pricing page could not be loaded for this update. ChatGPT prices come from OpenAI's pricing docs. The two $100 plans buy different things. ChatGPT Pro $100 buys 5x Plus usage. SuperGrok Plus buys more weekly usage, 1080p video, and priority access on the same model the $30 tier gets.

At the top, SuperGrok Heavy is $300 and ChatGPT Pro 20x is $200. As of September 10, 2026, OpenAI has paused new sign-ups and upgrades to the $200 Pro tier, so a new subscriber can only buy Pro $100 today. X Premium subscriptions also bundle Grok. Their prices are set by X, and we could not re-verify them this month.

Message Caps and Rate Limits (Weekly Usage Limits)

Both products dropped short-window message counts for their main tiers. ChatGPT made its cheapest model unlimited. Grok folded every product into one weekly budget.

Grok's pool is shared. A long Imagine video session eats into the same allowance as your chats and your Grok Build coding runs, because each product costs a different amount per action depending on compute. xAI's FAQ is blunt about top-ups: Extra Usage Credits are billed at standard rates, which costs more per action than the included pool, so frequent top-ups mean you are on the wrong tier.

ChatGPT's free-tier change is bigger than it looks. One Hacker News user summed up the old Free tier as about 5 messages a day on GPT-5.5 Instant, then unlimited GPT-5.5 mini (HN thread). Luna now covers all text chat with no count. The legacy GPT-5.5 Thinking limit on Plus and Business, up to 3,000 messages per week, still appears in OpenAI's GPT-5.5 help article for people who pick that model manually. GPT-5.5 leaves ChatGPT on October 14, 2026. The new caps sit on GPT-6 Pro instead. Pro $200 gets 200 GPT-6 Pro prompts per week, Pro $100 gets 50, and Business Premium gets 50 per week shared.

Model Lineup and Context: GPT-6 Sol, GPT-5.6 Sol vs Grok 4.7, Grok 4.6

OpenAI now ships three sizes per generation. Sol is the flagship, Terra sits in the middle, and Luna is the fast, cheap tier. GPT-5.6 used all three. GPT-6 has no Terra. GPT-6 Astra came first on September 3 and sits above Sol at $10 / $50 per 1M on the API. Sol and Luna followed on September 22. In ChatGPT, GPT-6 Sol and Luna are live in Work and Codex for Plus, Pro, Business, Enterprise, and Edu. Free and Go users get GPT-6 Luna in the desktop app.

xAI shipped Grok 4.6 on August 12 and Grok 4.7 on September 21, 2026. Grok 4.7 uses a larger base model than 4.6 and a longer RL run weighted toward multi-hour tasks, at the same price. Its context window is 500,000 tokens, half of what Grok 4.3 offers. If you need 1M tokens in one Grok request, Grok 4.3 and Grok 4.20 are the ones that still have it. Grok 4.7's knowledge cutoff is May 2026. xAI's docs now call the company SpaceXAI, and Cursor, which SpaceX finished acquiring on August 14, co-trains the Grok models.

Benchmarks: Grok 4.7 vs GPT-5.6 Sol

The only head-to-head published so far is xAI's own table in the Grok 4.7 announcement. It compares Grok 4.7 at xHigh effort against GPT-5.6 Sol at Max. GPT-6 Sol shipped a day later and is not in it.

Read the table with the source in mind. In the same chart, GPT-6 Astra scores 1,542 on GDPval against 1,695 for Grok 4.7, and a Hacker News commenter noted that GPT-5.6 Sol "performs suspiciously poorly" next to earlier models (HN). On a neutral composite, xAI's Grok 4.6 post cites Artificial Analysis: Grok 4.6 and GPT-5.6 Sol both score 61 on the AA Intelligence Index. OpenAI reports GPT-5.6 Sol at 53.6 on Agents' Last Exam and 80 on the AA Coding Agent Index at max reasoning.

Two numbers help fill the GPT-6 gap. OpenAI's GPT-6 Sol post reports 68.8% on DeepSWE v1.1 at max effort. xAI reports 71.0% for Grok 4.7 at high effort. Both are vendor-run, so treat the 2.2-point gap as noise. On arena.ai's Code Arena (September 22, 2026, human votes), gpt-6-astra-max ranks 1st at 1793 and grok-4.7-xhigh ranks 10th at 1632.

Feature Matrix

Grok Bot is the feature driving new "Grok bot pricing" searches. It launched in beta on August 11, 2026. Since August 26 it is included with SuperGrok, SuperGrok Plus, and SuperGrok Heavy, and with Cursor Pro, Pro+, Ultra, and Teams plans. It runs on its own usage, so work you hand to a bot does not draw down your Grok weekly pool. SuperGrok Lite is not on the list.

Pick by Task

API Pricing: Grok 4.7 vs GPT-6 Sol vs GPT-5.6 Sol

In June the API gap was 12x on output: Grok 4.3 at $2.50 versus GPT-5.5 at $30. The new flagships are much closer. Grok 4.7 matches GPT-6 Sol on input at $2 and undercuts it on output, $6 versus $10. GPT-6 Sol costs half of GPT-5.6 Sol.

Two billing rules change the math. First, Grok 4.7 has a 200K cliff. Once a prompt reaches 200K tokens, every token in that request bills at $4 / $12, including the first 200K. GPT-6 Sol's long-context premium starts at 272K. Second, cached input. Take an agent session that sends 1M input tokens in total, spread over requests that each stay under 200K, at a 90% cache hit rate, and gets 50K output tokens back. On Grok 4.7 that costs about $0.95. On GPT-6 Sol it costs about $0.88. If the uncached 10% is written to OpenAI's cache, it bills at the $2.50 cache-write rate instead of $2, and the total is about $0.93. The cache discount still outweighs Grok's cheaper output. Grok 4.7 also has no Batch API. OpenAI's Batch runs at 50% of standard price.

Rate limits on Grok 4.7 are 150 requests per second and 50M tokens per minute. Grok 4.7 Fast costs twice the token rates. It runs only in Cursor and Grok Build, not on the public xAI API.

// Morph Router returns a model choice. You call that model yourself.
const res = await fetch("https://api.morphllm.com/v1/router/multimodel", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.MORPH_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    input: userQuery,
    allowed_providers: ["openai"],
  }),
});
const { model } = await res.json();

// Easy prompts -> cheap tier (Luna)
// Hard prompts -> flagship (Sol)

Frequently Asked Questions