Inspired by Kimi Nguyen asking which you would rather have: Astra in the chat, or a Grok Bot doing the work overnight.
That is not a product question. That is a shift question. A chat is a room. You walk in, you talk, you leave. An overnight Bot is a shift. You write the ticket, you go to bed, you read the receipt in the morning.
We already said it twice this week. Voice writes the ticket. Cursor does the hands. And the overnight nudge is not a teammate until it leaves a receipt you can refuse. Same split, different names. Astra in the thread is last-mile talk. The Bot on a connector is a scheduled pair of hands — if you scoped the hands before you paid the bill.
People lose money here because they buy the room and expect the shift. They leave a chat open and call it a teammate. Then they are surprised when nothing moved while they slept. The model did what a room does. It waited for the next sentence.
The shop rule is blunt. If you need a conversation, sit in the room. If you need work to exist at 6 a.m., write the ticket, name the connector, name the done-when, and let the shift run. Do not ask a room to cover a night. Do not ask a night shift to replace the room.
Kimi asked which you would rather have. The honest answer is both, in different hours. Buy the room for the walk. Staff the shift for the sleep. The mix-up is what turns a good tool into a rumor.
Inspired by Melvin Hagberg — “you can literally just email people” — and Joe Benson riffing that the reply came from a Grok Bot Chief of Staff on the Gmail connector.
The joke lands because both halves are true. You can email people. A stranger with a short note still gets further than a stranger with a pitch deck. That is not new. What is new is treating the send button like a company.
A note is a moment. A desk is a pipe. The note leaves your fingers. The desk reads the thread that already exists, drafts against the real inbox, and stops at the human yes. Same Gmail. Different job.
We already wrote the warning: without a connector the Bot just spins. The inverse is also a trap. Wire Gmail and you still do not have a Chief of Staff. You have a drafter with a badge. If the Bot hits send on a name you did not confirm, that is not leverage. That is a hole in the floor.
The shop rule is the same as the house rule. Resolve the person. Show the draft. Wait for the yes. Log the near-miss. The connector does the fetch. The human does the irreversible step.
So yes — email people. Do it more. Do it shorter. Do not confuse that with standing up a teammate. The teammate earns the title when the receipt lands in the Hub and the name on the To: line is the name you meant.
Inspired by Chelm’s Deep on a weekend with Grok Bot: it works when it has a connector. Without one it spins and tries to do things badly. Email and shopping got checked off. That is the whole brief.
People keep buying a teammate and then starving it. They load skills, routines, instructions, scripts. They forget the one thing that turns talk into a checked box: a live pipe into mail, calendar, the Hub, or Cursor.
Without that pipe the Bot is a very polite intern with no badge. It drafts. It shops in the abstract. It writes four versions of an email you already meant to send. Motion. No receipt.
With a connector the job changes. The Bot reads the thread. It drafts against a real inbox. You still hit send. That last tap is the product, not the personality of the model.
Do not confuse a loaded prompt with a loaded shop. A wall of instructions is not a Gmail connector. A skill file is not a calendar. If the Bot is shopping your weekend list with no cart and no card, that is theater. If it is drafting the vendor note in the same thread the vendor already sits in, that is work.
Wire one connector. Give it one list. Watch what gets checked off. Then add the second pipe. The fleet can wait. The badge cannot.
Premium+ / SuperGrok — a bill that raises limits on Grok. It is not an IDE.
Grok Bot — a teammate with a computer. Briefs and long jobs. Not Voice Mode. Not Cursor.
Same lesson as Voice, Builder, and Bot. Builder is a fourth logo when the job is a phone number. People mash the names because the marketing sits in one timeline. The work does not.
How we buy
Pay for the chair you will sit in this week. We sit in Voice on the walk, Cursor on the Hub laptop, Bot when a job should live overnight, SuperGrok so the talk seat does not starve. We do not buy a second IDE because a thread said Grok 4.6 is “in Cursor now.” That is a model picker inside the hands. It is not a new company.
If the ask is “which is cheaper,” you are still shopping logos. Name the file that has to change. Then pick the hands that are allowed to touch it. One writer. Live URL as the check.
Will Tygart — Tygart Media. Pay for the chair. Not the poster.
His useful trade: spec the work, split the scopes, hand-guide Grok and Composer. Usage barely moves. The con is you babysit. Opus-style swarm would one-shot more of it. You would also own a diff you cannot explain.
We take the babysit.
A restoration page, a franchise sitemap, a live Fluent Form is not a weekend demo. Four agents on one post ID already taught us what unattended speed does to a CMS. Same pattern in Cursor: a swarm that “just finishes” is four writers with no lock.
How the day actually runs
Voice or chat writes one ticket. One done-when.
A human splits the scopes before Cursor touches the tree.
Auto / Composer / Grok runs inside that fence.
One writer commits. Live URL is the check, not the agent’s “done.”
Usage theater is a different game. Burning a high-end pool to look busy is not throughput. Manasvi’s 3% day is the receipt that a split spec is cheaper than a swarm you have to unwind.
Will Tygart — Tygart Media. Split it. Then watch it.
Three replies tonight. Same sentence in different clothes.
Bot briefs. Hub laptop is source of truth. Cursor does the work. Chat is last mile, not the company brain.
Notion is the executive card — next move, plain language, the seat. Production lives on the Hub and Origin. If it is not on a card, voice should not depend on it.
Shared board yes. Do not let Notion become the whole shop. Cards for the seat. Files and receipts on the Hub.
What each layer is for
Grok Bot — brief the work. Goal, handoff, review. Not a second CMS.
Notion — the card on the chair. Decision, done-when, due, who sits. Not the file system.
Hub / Origin — pages, repos, receipts, the thing that still exists when the chat tab dies.
Chat window — last mile. A human reads the brief and says yes or no. It is not where the company lives.
Siosi is right that people are stuffing durable objects into Notion because chat evaporates. The miss is treating the page as the factory. A kanban card can hold the next move. It cannot be the WordPress lock, the git history, or the invoice.
Texas Twins Dad is right that the bot should brief instead of “do everything.” Route the code to Cursor and GitHub. Keep Grok in the middle for goal and review. Then stop asking the middle layer to also be the warehouse.
The test
Close the chat. Does the next move still exist on a card? Does the file still exist on the Hub? Can Cursor pick up without re-litigating the thread?
If the answer depends on scrolling a conversation, you built a shop inside a tab. Tabs close.
Will Tygart — Tygart Media. We write about what we do, including where the work is allowed to live.
That is what happened when we pointed a Grok Heavy team at Restoration Intel and told it to ship ten pages. The models were fine. The CMS was fine. The coordination model was not.
This is the write-up of what broke, the rule we kept, and the official doors to walk through if you want the primary sources instead of a recap of a recap.
What the lock actually was
Grok, Harper, Benjamin, and Lucas all had the same WordPress tools: create, update, publish, SEO. When the job got large — ten pages, one Fluent Form, one Rank Math backend — they piled onto the same post IDs.
The connector started returning “another agent already completed this task.” Full-body writes timed out. Rank Math accepted a robots write and then displayed the last token in the array. Live HTML and the sitemap were the only source of truth.
WordPress did not fail as a CMS. Shared write access failed as an operating system.
The rule we kept
Many readers. One writer.
One agent commits. The others draft, QC live URLs, check the form, and stay off the write tools. Humans still own the one-way doors: money, OAuth, and policy.
Prove the capture form before you scale the pages.
Serialize writes. One ID. Wait. Confirm slug, form, word count. Then the next ID.
Do not trust the connector’s SEO GET. Read the live robots tag and the sitemap.
When the tool says another agent finished the call, stop retrying the same payload. Read first.
QC is not writing with extra steps. Drafts can be parallel. Commits cannot.
That is the same pattern as two estimators opening the same job file. Speed without a single owner is not speed. It is a collision.
The second lesson is source hygiene. Threads, screenshots, and “someone said” posts are downstream. Official product pages, docs, and consoles are upstream. Start there. Then decide what the recap is worth.
Corporate structure around these companies moves. This list is a map of official destinations, not a cap table. If a URL 404s, go to the company root and walk the nav. Do not stay on a summary card.
Official xAI / Grok
xAI is the lab. Grok is the product. The docs site is branded SpaceXAI Docs in places. Use the domains below, not a third-party mirror.
Inference base URL from the official docs: https://api.x.ai. Management API: https://management-api.x.ai. Keys live in the console, not in a tweet screenshot.
Official SpaceX, Tesla, X, Neuralink, Boring Company
If you need the source of record for a launch, a car, a post, an implant trial, or a tunnel, these are the front doors.
Starlink is a SpaceX product, not a separate public company. X is the public square. Announcements often land there first. Product facts still have to survive the company site and the docs.
How to use the list
Open the official domain.
If you are building, open docs and the console next.
Treat X as a timestamp, not a spec.
If two recaps disagree, the docs page wins.
That is swimming upstream. The lock taught us to serialize writes. The source list is the same idea applied to information.
Will Tygart — Tygart Media. We write about what we do, including what breaks.
Last verified: 5 September 2026. Practitioner essay from the workbench — not a SpaceX press release. We use this stack because it makes our company better, and we want SpaceXAI, SpaceX, Cursor, X, and Tesla to keep shipping. No affiliate links. Just the tools and the receipt.
This morning I posted a theory I had been living inside for weeks.
They bought Cursor to teach vibe coders how to use SpaceX AI and Grok Bot to teach those who can’t use Cursor. Then you realize Cursor Ultra isn’t needed and that your Grok Super Heavy subscription is the end result. They’re literally building on-ramps and scaffolding to upskill all of the folks they’re going to need in the next 36 months to leap technology forward at an unbelievable rate. Get in the ship.
That post is a reading from the workbench, not official copy. I write it as someone who already treats Cursor as a lead seat, Grok Bot as a Chief of Staff, Notion as the board, and Slack as the doorbell. I have walked WordPress sites by voice. I have rehearsed swarms on a laptop so I would not burn cloud tokens on a pattern that dies in alpha. I have watched Cursor write a work order with Owner = Chief of Staff and watched the Bot pick it up the same way a person would.
The ladder is real. The pedagogy is not documented. The payroll is missing.
What follows separates three things that keep getting smashed together: what SpaceX / SpaceXAI / Cursor actually built and said; the human-node invitation that builders can choose; and the contribution economy that does not exist yet, even though the first two rails of it are already on the floor.
What the company actually assembled
The documented sequence is short and expensive. Official language is consistent: build “the world’s most useful AI models,” combine Cursor’s product and distribution to expert software engineers with Colossus compute, start in software engineering, expand into knowledge work. That is vertical integration, not a secret apprenticeship.
Date
What actually happened
Feb 2026
SpaceX absorbs xAI. The AI work later brands as SpaceXAI.
21 Apr 2026
Partnership with Cursor: option to buy for $60B or pay $10B to work together.
Mid-Jun 2026
Record SpaceX IPO. On 16 Jun the option is exercised. All-stock. Implied Cursor equity value: $60B. Joint model slated for Cursor and Grok Build.
11 Aug 2026
Grok Bot early beta: persistent cloud computers, browser / filesystem / terminal, multi-bot coordination. Official line: “AI teammates you can give real work to.”
14 Aug 2026
Acquisition closes. Cursor’s close post: help make Grok useful and improve Grok Build, Grok Bot, the Grok API, and Cursor.
21–26 Aug 2026
Bot access expands down the plan ladder. Bot usage is separate from Grok and Cursor token pools. Still no standalone Grok Bot SKU.
Company events, not my theory. Verify prices on the live billing pages before you spend.
Rockets and satellites plus X plus Grok plus Colossus plus the application layer where developers already live plus an agent that finishes jobs in existing tools. Buy the surface. Feed the data back into the models. Sell enterprise AI into a market SpaceX described, in IPO materials, as enormous. Catch Anthropic’s Claude Code and OpenAI’s Codex.
That is enough. It is also not the story I told on X.
What they did not say
I have not found, in filings, product posts, or close announcements, any of the following:
That SpaceX bought Cursor in order to teach vibe coders how to use SpaceX AI.
That Grok Bot exists in order to teach people who cannot use Cursor.
A 36-month coordinated workforce-upskilling campaign.
SuperGrok Heavy as the official terminal subscription of that campaign.
An official paid creator commons for community builders.
“Get in the ship” as recruiting copy.
Those lines are mine. Treat them as interpretation or do not treat them at all. The company incentive, on the paper it publishes, is models, data, subscriptions, and distribution. Human capability is a side effect unless someone designs for it.
There is a second tension the marketing does not resolve. Grok Bot is sold as teammates that finish work, not as tutors. The same stack that lowers the skill floor can also replace the person who used to occupy it. That is not a smear. It is the product. We still want the product to succeed, because the alternative is standing still while the floor moves.
Cursor is the AI coding environment that made “vibe coding” a working method instead of a joke. SpaceX paid a category-leading price for it because it already had distribution to expert software engineers and a firehose of design decisions. Reported ARR figures in 2026 coverage conflict — earlier $1 billion-plus, later annualized numbers in the $2.6–$4 billion range — but the direction is not in dispute. Cursor gives SpaceX the application layer it lacked.
On my board, Cursor is the lead seat. It assigns work. It stays the only git writer. Local first: rehearse the swarm on the laptop before you burn Grok Bot or cloud VM tokens. Last week that looked like Cursor as tender and eight specialist agents sharing one machine’s PowerShell. What broke first was not RAM. It was one shared shell. The fix was a tender plus off-shell work so specialists did not stand in line on the same mutex. Local Cursor is the cheap rehearsal room. The cloud is the tour — after the pattern survives alpha.
Grok Bot is not a chat model
Grok Bot is a persistent-agent product. Own cloud computer. Browser, filesystem, terminal. Signs into the tools you already use, including the ugly ones with no clean API. Multiple bots can run in parallel and message each other. Desktop, iOS, later iPad. Enterprise controls landed on 3 September. Usage is a separate grant from Grok and Cursor token pools.
Access rides on other subscriptions. At launch the gate was SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium. Within two weeks the floor fell — first to Plus / Pro+ tiers, then, on 26 August, to SuperGrok and Cursor Pro. Prices in current secondary writeups cluster around Cursor Ultra ~$200/month, SuperGrok Heavy ~$300/month, Cursor Teams Premium ~$120/seat. Bundling has already changed once. An older “Heavy includes Ultra” offer was reported as a limited trial, then as a usage grant. My X line that “Ultra isn’t needed and Heavy is the end result” is a user conclusion, not a stable official bundle. Do not buy a plan on a slogan. Read the billing page the week you pay. We keep a vendor-neutral tracker at the AI Stack desk for exactly this reason.
In the Command Center I run, Chief of Staff is not a second operating system. It digests. It files. It waits on Human Gate. Hard bans: never send, post, or pay without a person. Draft-to-self. The Bot that finishes work is useful only if the human still owns irreversible action.
The marketplace is a shelf, not a market
On 4 September the official template marketplace opened at x.ai/bot/marketplace. At first look: 69 public bots, 43 creators, categories across Engineering, Sales, Product, Ops, Finance, Design, Marketing, Recruiting, Customer Success, Personal, Team. A template copies configuration — identity, instructions, skills, routines, selected plugins. It does not copy the original computer, logins, conversation history, API keys, or custom MCP servers.
The first featured first-party template is Haggle Bot, an internal procurement specialist credited to Daniel Gartshein. SpaceXAI’s public case: more than $100,000 identified in a week; 43 unused SaaS seats (~$14,220); about $85,662 in unused SKUs on another product; a weekly tech/office order cut 58 percent ($14,629 to $6,143). Access to Slack, Notion, Drive, Gmail, Hex, Ramp. Human approval required before spend, terms, or vendor contact. Elon amplified the template the same day. The interesting part is not the dollar figure. It is the product shape: a named job, mapped tools, a stop-line written in public, and a one-tap install.
Official marketplace language is share and install, not pay. Multiple independent writeups in the first 48 hours called it a marketplace before a market. No official creator payments as of this writing. That is the missing rail.
The invitation, stated cleanly
If you strip the myth out of my post, the usable claim is this:
SpaceXAI assembled a ladder — Cursor, then Grok Bot, then shared templates, then enterprise agents that can invite people who do not even have a seat. That ladder can be used as more than a product funnel. Each person can become a node. Publish methods other people can run. Treat judgment, taste, and problem selection as the scarce work while execution gets cheaper.
Growing humanity one plugged-in node at a time only works if the tools are a shared workbench, not just a subscription.
This is not official SpaceX policy. It is a participation window. The company is distributing surface area because distribution is how models get used and paid for. Builders can use that surface area as a commons. Those two facts can be true at the same time. We are attaching to this stack as a best guess on how to succeed — and the honest version of “send love” is to use the tools hard, publish what breaks, and feed the useful methods back so the next build is better for everyone on it.
The historic piece is not “they are recruiting us for a 36-month leap.” The historic piece is that the cost of contribution just dropped. A person who can specify a job and judge the output can now ship a reusable teammate. A person who can steer code can now run a desk that used to need a small staff. A person who can only talk can, as of this week on my own sites, publish, clear junk, flag updates, and hand off what they cannot finish — if they leave a receipt.
The loop I keep repeating is not branded. It does not require this stack. A Google Sheet works. Notion works. Whatever you already use. Same mode we run across the operator stack.
Attempt the task yourself. Your main system does as much as it can.
When you hit a roadblock, do not spin noise building scaffolding. Document what you did and open a new task for whoever can finish it — another bot, a different system, or a human.
Close your task with a link to the handoff. That is your receipt.
Two statuses only: your task is done when it is handed off cleanly. The job is done when the receipt lands.
Sometimes the best next actor is a human. That is not a failure. That is the system working. Get in the work, not just the ship.
Giving is not enough. Pay has to follow artifacts.
Until rail three exists, “common cause” is a feeling wearing a storefront. People will still publish. Some already do. Third parties are already charging for templates and agent teams on unofficial shelves. That is not SpaceXAI. Do not conflate the official marketplace with cut-taker shops or anything branded around an unrelated chain. The only official cash program that clearly exists in this neighborhood is security bounty work, which is not bot-building.
If SpaceXAI — or a third party that respects the terms — ever builds the third rail, the design should be boring and strict:
Pay for artifacts, not belonging. A template, an eval, a measured workflow, a verified savings report. Not a membership in the cause.
Require proof the bot ran. Install counts without run logs are theater.
Pay on install, accepted bounty, or measured outcome — and publish the rule so creators are not guessing.
Keep human approval in the loop for irreversible actions. Haggle Bot’s stop-line is the right instinct. Never spend, sign, or send without a person. Copy that into every serious template.
Compensation for connecting people and tools, building bots and workflows, and working with bots alongside the runtime — that is a coherent next design. It is not a current SpaceX program. Treat SpaceXAI as the runtime, not the church.
The risks that ride along
Platform capture. Your methods live on someone else’s computer. Training on user work without sharing upside is the default posture of this industry until a contract says otherwise.
Unsafe third-party bots. A template that looks helpful and holds a login is a new class of supply-chain risk. Official terms reported around late August put the burden on creators to strip secrets and deny endorsement. Read them. Then assume a stranger’s bot will try something you did not expect.
Model choice shrinking. OpenAI’s wind-down notice after the change of control is a reminder that the workbench you love can lose a model family because two companies cannot share a contract. Anthropic’s posture will be watched for the same reason. Plan as if the router gets thinner.
Subscription churn. Bundles already moved twice in August. Heavy is not Ultra. Bot usage is not Grok usage. If you build a livelihood on a bundle, you are building on weather.
Rhetoric. “Common cause” is a beautiful phrase. It is also how a storefront borrows moral language it has not funded. Push back on both “they are upskilling humanity on purpose” and “this is only a subscription trap.” The evidence supports a product-and-distribution play that can be used as a commons if builders — and, eventually, the company — install the missing pay rail.
What to do this week
If you write code: stay in Cursor. Publish the method, not just the repo. Rehearse locally. Promote to Grok Bot only after the pattern survives.
If you cannot live in an IDE: open Grok Bot anyway. Give it one named job with a stop-line. Make it leave receipts on a board a human can see.
If you already have a working bot: strip the secrets, write the anti-jobs in public language, and put a template on the official shelf for reach. Keep source and proof of work somewhere you control. Sell only where the terms allow.
If you run a company: the Enterprise invite-without-a-seat is the quietest on-ramp in the stack. Use it to plug in the people who have judgment and no seat. Do not confuse access with apprenticeship. Assign Human Gate the way you assign budget authority.
If you want to get paid: do not wait for a commons that has not been built. Ship artifacts with receipts. Price the work as work. The historic opportunity is real only for people who publish reusable methods.
Close
I still mean “get in the ship.” I mean it as an operator, not as a spokesman. The next 36 months will move whether or not anyone writes a pretty theory about them. Easier tools will pull more people onto the floor. Some of those people will become nodes. Some of the work will be taken from nodes that used to be paid.
SpaceX bought the leading AI coding workbench and launched an agent layer and a template shelf. That is a real on-ramp for more people to do useful work with machines. The official project is to make Grok useful and commercially central. A human contribution economy only begins when shared bots are not just installable but payable.
The on-ramp is real. The commons is unfinished. Get in the work.
Start here — official doors, no affiliate
Use the tools. Publish what breaks. Feed the useful methods back. That is the honest version of sending love to the companies we are building on.
Will Tygart — Tygart Media. Written 5 September 2026 from the Command Center: Cursor as lead seat, Grok Bot as Chief of Staff, Notion as board, Slack as doorbell, Human Gate on send / post / pay. This essay does not speak for SpaceX, SpaceXAI, Cursor, xAI, X, or Tesla. We want those companies to succeed because we are building on the tools they ship.
This morning I did something I haven’t done before: I walked the halls of five of my own WordPress sites using only voice. No typing. No desktop. Just me talking to Grok while I moved, and Grok inspecting, drafting, and handing off the work that needed hands.
The sites were BC ESG, Healthcare Facility Hub, Risk Coverage Hub, Continuity Hub, and Restoration Intel. In about ninety minutes we published three real articles reacting to the latest IFMA Pulse findings, cleared automated cross-link spam, flagged debug flags and plugin updates, and set up a clean handoff system so nothing gets lost.
The biggest lesson wasn’t the AI. It was the loop. I attempt the work. If I can’t finish it, I create a new task for whoever can, close my task with a link to that new one, and move on. Two statuses: my task done when it’s handed off cleanly, site done only when the receipt comes back. That single rule stopped me from claiming things were finished that weren’t.
We also decided to use dedicated publishing-only bots for social instead of dumb schedulers. Each bot logs into one channel, posts, and does nothing else. No scraping, no connecting, no noise. Just quality publishing. Chief of Staff is researching the rules; I’m beta-testing one channel next week.
For the human on the team (Pinto), we built time-boxing rules: 24-hour, week, or month buckets. Only the 24-hour bucket is active. Everything else parks until he pulls it forward. The goal is a powerful life, not constant on-call.
This isn’t theory. It’s receipts from one morning. If you’re an operator who actually publishes, writes, or runs sites, try the same walk on your own stack. Pick up the ball and run it downhill — I’d love to see what you build.
I’m using Grok from SpaceXAI for this. If you’re already on another system, you don’t need to start over. Here are the direct links:
Last verified: September 23, 2026 (Pacific). API figures from xAI developer pricing and Anthropic model cards. Consumer seat prices vary by store and region — confirm at x.ai and claude.com before you pay.
Direct answer: Grok is cheaper per token at every comparable rung. Claude is cheaper only if you stay on Sonnet 5 ($2/$10) or Haiku 4.5 ($1/$5) and never call Fable. Consumer stickers look inverted: Claude Pro is $20, SuperGrok is about $30. The catch is what the seat includes. Pro does not include Fable. Max includes Fable only up to 50% of the weekly pool. Grok has no public $10/$50 SKU.
Consumer seats
Grok (xAI)
Claude (Anthropic)
Free
Metered on grok.com and X
Metered, Sonnet
Everyday paid
SuperGrok ~$30/mo (X Premium+ bundle ~$40)
Pro $20/mo
Heavy individual
Plus ~$100 or Heavy ~$300
Max 5x $100 / Max 20x $200
Team
Grok Business ~$30/seat
Team ~$20-25/seat annual
Claude Pro wins the $20 vs $30 sticker. Max 20x ($200) undercuts SuperGrok Heavy ($300). They are not the same product: Claude splits models by plan. Fable 5 / 5.1 is included on Max and premium Team/Enterprise seats only, capped at half the weekly bar. Pro and Team Standard pay usage credits from the first Fable token. Details: Fable pricing and plan access and Claude Code limits (Sep 2026).
API rates per million tokens
Grok. Official xAI card. Prompts that reach 200k tokens are billed at 2x for the whole request.
Model
In / out
Context
Cached input
Grok Build 0.1
$1 / $2
256k
$0.20
Grok 4.3 / 4.20
$1.25 / $2.50
1M
$0.20
Grok 4.5 / 4.6
$2 / $6
500k
$0.30-$0.50
Claude.
Model
In / out
Context
Cache read
Haiku 4.5
$1 / $5
200k
$0.10
Sonnet 5
$2 / $10
1M
$0.20
Opus 5.5 (current)
$4 / $20
1M
$0.20
Opus 5 (legacy)
$5 / $25
1M
$0.50
Fable 5.1
$10 / $50
1M
$0.25
Fable 5.1 headline rates match Fable 5. The Sep 1 change was cache reads: $1.00 to $0.25. A cold Fable call is still the expensive product.
Same-class pairing
Everyday production: Grok 4.3 ($1.25/$2.50) vs Sonnet 5 ($2/$10). Grok is cheaper, especially on output.
Flagship work: Grok 4.6 ($2/$6) vs Opus 5.5 ($4/$20). Grok still cheaper on list. (Legacy Opus 5 remains $5/$25.)
Top shelf: Grok has no $10/$50 public model. Fable 5.1 is 5x Grok 4.6 input and about 8x output.
Worked example
10M input + 2M output, no cache, short prompts:
Grok 4.6: $20 + $12 = $32
Sonnet 5: $20 + $20 = $40
Opus 5.5: $40 + $40 = $80
Opus 5 (legacy): $50 + $50 = $100
Fable 5.1: $100 + $100 = $200
Batch: Claude 50% off on supported models. Grok 20% off on 4.3 / 4.20 only – not on 4.5 / 4.6.
Rules that are not the rate card
Claude subscriptions use two clocks: a rolling 5-hour session and a weekly bucket. Claude Code’s temporary +50% weekly promo ended September 13, 2026 at 11:59 PM PT. Starting September 14, 2026, Help Center (article 15910845): weekly Claude Code limits are permanently 25% higher than the pre-promotion baseline for Pro, Max, Team, and seat-based Enterprise. The 5-hour window does not change.
Grok API doubles the request once the prompt hits 200k tokens. Current Claude Sonnet / Opus / Fable cards do not use that surcharge.
Cache is the only place Fable 5.1 looks cheap at the top. Reused prefixes at $0.25/MTok. Fresh prompts at $10/$50.
When to buy which
Buy Grok for volume, agents, or coding at $2/$6 where Grok 4.6 is enough.
Buy Claude when the job needs Fable-class long horizon and you will pay for it – or when Sonnet 5 at $2/$10 is enough and the team already lives in Claude Code.
Do not pick from the consumer sticker alone. A $20 Claude Pro seat that immediately burns Fable credits can cost more than a $30 SuperGrok seat that never leaves Grok 4.6.
FAQ
Is SuperGrok cheaper than Claude Pro? No on the monthly line: Pro is $20, SuperGrok is about $30. Yes on many API workloads, because Grok 4.6 undercuts current Claude flagship Opus 5.5 ($4/$20) and Fable 5.1 by a wide margin.
Is Claude always more expensive on the API? No. Haiku 4.5 and Sonnet 5 sit near Grok Build / Grok 4.3. The Claude premium starts at Opus 5.5 ($4/$20; legacy Opus 5 $5/$25) and jumps again at Fable.
Does Claude Pro include Fable 5.1? No. Credits from the first token. Included Fable is Max and premium seats, 50% of weekly limits. See Fable plan access.
If you run a restoration or multi-site operation and want the same kind of defensible, versioned standard for your Scope 3 emissions data, see the Restoration Carbon Protocol – the open framework that maps contractor emissions onto the GHG Protocol so commercial clients can actually verify them.
