Make Review September 2026: The ChatGPT Plugin, Claude Fable 5.1, and What October 1 Changes

Make shipped three buyer-relevant changes since our August review rated it 4.5: Claude Fable 5.1 joined the AI modules on September 4, Global Search landed August 31, and on September 9 the official Make plugin for ChatGPT went live (help.make.com/2026, retrieved 2026-09-10; community.make.com, retrieved 2026-09-10). The rating holds at 4.5. All three are additive, and none of them touches the two structural cons that keep Make under a 4.7.

Here is the one thing to leave with. August's headline was Maia, an AI co-worker living inside Make's Scenario Builder. September runs it the other way around: Make now works inside ChatGPT, where you can build automations, create AI agents, and browse execution history without opening Make at all (community.make.com, retrieved 2026-09-10). That flips which window is primary for a large slice of users. It does not retire Maia. The plugin is for describing and checking work from a chat you already have open; Maia in the Scenario Builder still owns the deep building and editing you do with the modules laid out on the canvas.

What changed since August 21

Four dated items carry the delta. August's review is the baseline for everything below.

Axis	August 2026 baseline	September 2026	Source
Primary AI interface	Maia inside the Scenario Builder	Plus the Make plugin inside ChatGPT, ChatGPT Work, and Codex (Sep 9)	community.make.com
Top-end model in AI modules	Claude Sonnet 5, Opus 5	Plus Claude Fable 5.1 at 10 / 50 USD per MTok (Sep 4)	help.make.com/2026
Cheap, fast model	Gemini 3.7 Flash, 0.75 / 3.75 USD intro	Unchanged	Aug review
Search	Per-area search	Global Search across scenarios and runs (Aug 31)	help.make.com/2026
Cycles per run	Present, removal announced	Removed October 1, 2026	help.make.com/2026
Pricing	Core 9 USD / 10,000 credits	Held	make.com/en/pricing

Global Search is the minor one: since August 31 you can find scenarios and scenario runs from a single box (help.make.com/2026, retrieved 2026-09-10). Useful when you manage dozens of scenarios, not a reason on its own to rerun this review. The plugin and Fable 5.1 are.

The Make plugin for ChatGPT: when to use it, when to open Maia

The plugin lists in OpenAI's official ChatGPT plugin directory, and you sign in with your existing Make credentials (community.make.com, retrieved 2026-09-10). Inside a ChatGPT conversation you can build automations from a plain-language description, create AI agents in the same chat, search your automations by name, folder, or status, and review execution history with searchable tables, status badges, pagination, and structured results. It runs in the browser, in the ChatGPT Work desktop app, and in Codex, and it is available on all Make plans, including Free.

One line in Make's announcement is the load-bearing detail: "Your automations continue running in Make's cloud whether or not ChatGPT is open, with your existing connections, permissions, scheduling, retries, and error handling" (community.make.com, retrieved 2026-09-10). The plugin is a control surface over the same cloud engine, not a second runtime. Your permission model and connections carry over untouched.

So who lives where? Use the ChatGPT plugin when your work already runs through ChatGPT: you want to spin up a quick automation from a description, check whether last night's run succeeded, or stand up an agent mid-conversation without context-switching to another tab. Open the Scenario Builder and use Maia when you need to see the modules, wire a 30-step flow with branching and error handlers, or edit a specific module's mapping by hand. The plugin wins on speed and staying in one window; the builder wins on depth and precision. Neither is a replacement, and treating the plugin as a full builder is the mistake to avoid.

Claude Fable 5.1 vs Gemini 3.7 Flash in the AI modules

Claude Fable 5.1 arrived in Make's AI modules on September 4 (help.make.com/2026, retrieved 2026-09-10). It lists at 10 USD per million input tokens and 50 USD per million output, with cache reads at 0.25 (platform.claude.com, retrieved 2026-09-10). Set that against the two models August priced: Gemini 3.7 Flash at an intro 0.75 / 3.75 USD per million through December 31, and Claude Sonnet 5 at 2 / 10 USD. Fable 5.1 runs roughly 13x Gemini's intro input price and 5x Sonnet 5's. It is the premium reasoning tier of the dropdown, not the cheap-and-fast pick.

Anthropic points Fable 5.1 at demanding reasoning and long-horizon agentic work, and recommends starting with Opus 5, reaching for Fable 5.1 only when Opus 5 at higher effort comes up short (platform.claude.com, retrieved 2026-09-10).

That positioning has a direct credit cost inside Make. Every model call in a Make AI Agent draws credits based on the tokens the model burns (Aug review), so a Fable 5.1 step draws far more credits per run than the same step on Gemini 3.7 Flash or Sonnet 5. The pick, by task: reserve Fable 5.1 for demanding reasoning or long-horizon agent steps where the cheaper models genuinely fall short, and keep high-volume, simple agent runs on Gemini 3.7 Flash. Switching a busy scenario to Fable 5.1 for its own sake is an easy way to burn a month of credits on work Gemini would have finished.

October 1: the cycles-per-run removal

Make removes the Cycles per run setting on October 1, 2026, a change it announced August 13 (help.make.com/2026, retrieved 2026-09-10). August flagged this as a con, and with the date now about three weeks out, September is the last clean window to baseline before it lands. Cycles per run controlled how a scenario grouped its processing cycles inside a single execution, so looping and retry scenarios that leaned on it will meter credits differently once it is gone.

Make published no fixed dollar exposure, which makes this an audit item rather than a number. Open each looping or retry scenario now, note its credit-per-run baseline from the run history, then rerun after October 1 and confirm the meter reads the same. A scenario that jumps from 200 credits a run to 600 is something you want to catch in a test, not on the invoice.

Pricing, retrieved today

Make's plans held since August. Every price below is per month for the base 10,000-credit tier; the pricing page slider raises credit volume at adjusted rates (make.com/en/pricing, retrieved 2026-09-10).

Plan	Price / month	Credits / month	Source
Free	0 USD	1,000	make.com/en/pricing, 2026-09-10
Core	9 USD	10,000	make.com/en/pricing, 2026-09-10
Pro	16 USD	10,000	make.com/en/pricing, 2026-09-10
Teams	29 USD	10,000	make.com/en/pricing, 2026-09-10
Enterprise	Custom	Custom	make.com/en/pricing, 2026-09-10

The verdict, by who you are

Make earns 4.5 in September 2026. The ChatGPT plugin is the most interesting structural move since Maia, but it is a new front door onto the same engine, and the credit meter plus the self-hosting gap are untouched. The picks, by situation:

• Solo non-developer operator. Make, still, and September gives you two ways in. Describe a workflow to Maia inside the builder, or build and check it from the ChatGPT plugin without leaving the chat you already work in (community.make.com, retrieved 2026-09-10). Start on Free's 1,000 credits, then move to Core at 9 USD a month once a real scenario works (make.com/en/pricing, retrieved 2026-09-10). If you land here, Make's paid plans are the buy.
• Mixed-technical ops team. Make edges n8n. Private Spaces gives per-member credit caps and custom roles let an admin define precise organization and team permissions, all without a server to patch (help.make.com/2026, retrieved 2026-09-10). Teams runs 29 USD a month for 10,000 credits (make.com/en/pricing, retrieved 2026-09-10). The flip condition holds: if every member is technical and the team wants zero per-run cost more than managed governance, self-hosted n8n wins.
• Technical team that can self-host. n8n still wins on cost, and nothing in September moves that pick. Make added no self-hosting, so the per-credit meter stays, and neither the ChatGPT plugin nor Fable 5.1 changes that, since both point at your own model keys either way. Once monthly operations clear roughly 30 to 40k and someone can run a Docker box, self-hosting deletes the meter Make charges on. The cost that math never prints is ops time; if nobody wants the pager, Make's managed model is still rational even where the dollars favor n8n.
• Enterprise that needs custom roles. Make holds. Custom roles from August plus the plugin's respect for your existing permissions and connections (community.make.com, retrieved 2026-09-10) let non-builder staff drive automations from ChatGPT without loosening governance. The flip: if a compliance mandate requires the workflow engine to run on your own infrastructure, self-hosted n8n is the answer Make cannot give.

FAQ

Did Make's rating change in September? No, it holds at 4.5. The ChatGPT plugin, Claude Fable 5.1, and Global Search are all additive, but none removes the two structural cons (no self-hosting, credit fan-out at volume) that keep Make under a 4.7, and the plugin changes how you reach Make, not what Make can automate.

What is the Make plugin for ChatGPT? An official plugin, listed in OpenAI's ChatGPT plugin directory since September 9, that lets you build automations, create AI agents, and review execution history from inside ChatGPT, ChatGPT Work, and Codex, on all plans including Free (community.make.com, retrieved 2026-09-10). Your automations keep running in Make's cloud with your existing connections and permissions whether or not ChatGPT is open.

When should I use the plugin instead of Maia? Use the plugin when your work runs through ChatGPT and you want a quick build, an agent, or a run-status check without switching tabs. Open the Scenario Builder and use Maia when you need to see and edit the modules on the canvas for a deep or branching flow. The plugin wins on speed; the builder wins on depth.

Should I use Claude Fable 5.1 in my Make agents? Only for the hardest steps. Fable 5.1 lists at 10 / 50 USD per million tokens (platform.claude.com, retrieved 2026-09-10), roughly 13x Gemini 3.7 Flash's intro input price and 5x Sonnet 5's, and every model call draws credits by the tokens it burns. Reserve it for demanding reasoning or long-horizon agent work and keep high-volume, simple runs on Gemini 3.7 Flash.

What changes on October 1? Make removes the Cycles per run setting (help.make.com/2026, retrieved 2026-09-10), which changes how looping and retry scenarios group and meter their processing cycles. Make published no dollar exposure, so baseline your looping scenarios' credit-per-run now and rerun them after the change to confirm the meter reads the same.