Model freshness audit · By Adam Johnstone · Updated 2026-09-13
A bundle is a wrapper around someone else’s models, so the only question that matters is which models, at what size, updated how often. Nobody in this category audits that. This page does, with a date on it, because “access to leading AI models” is a marketing phrase and a model picker is a fact.
The audit
Checked 13 September 2026 against each vendor’s own pages. This is re-run and re-dated rather than written once.
| Tool | Models advertised | Generation | Do the vendor’s own pages agree? | Shipping activity |
|---|---|---|---|---|
| ChatHub | GPT-6 Astra, Claude Opus 5, Gemini 3.1 Pro | Current | Model page current; pricing page still says GPT-5 / Claude 4.5; docs still say GPT-4o | Docs changelog frozen since 18 Nov 2025 |
| Sider | GPT-5.5, Claude Opus 4.6, Gemini 3.1 Pro | Current | Homepage and credit-rules page agree | Five dated releases in 2026, latest 1 Sep |
| MultipleChat | GPT-6 Astra, Claude Fable 5.1, Gemini 3.8 Flash | Current generation, but Flash-class Gemini | Homepage says Gemini 3.8 Flash; subscription page says 3.7 | No public changelog |
| Monica | Store: GPT-5.2, Claude 4.5 Sonnet, Gemini 3 Pro. Homepage: GPT-5.5, Claude 4.7 Opus | Current, but sources disagree | Three official surfaces name three different sets; help centre still lists GPT-4o and Claude 3.5 | Extension updated 6 Aug 2026 |
| Merlin | o3, GPT 4.1, Claude 3.7 Sonnet, Gemini 2.5, Grok 3 | Two generations behind | Query-cost table prices only retired models and carries no date | One changelog entry in all of 2026 |
| TypingMind | Whatever your API key reaches | You control it | Changelog lists model additions within days of release | Latest entry 11 Sep 2026 |
| You.com | Not disclosed | Not applicable | APIs internalise model choice; provider not published | API launches through 2026 |
“Generation” compares what the vendor advertises against the current frontier releases. A tool is not marked behind for offering cheaper models — only for advertising nothing current. Sources are on each tool’s page.
Why this is the check that matters
Its site names o3, GPT 4.1, Claude 3.7 Sonnet and Gemini 2.5 — no GPT-5, no Claude 4, no Gemini 3 anywhere. Its official query-cost table prices only retired models. One third-party page claims GPT-5.2 and Claude Opus 4.6, and nothing Merlin publishes supports that. Open the model picker before you pay; it takes thirty seconds.
MultipleChat advertises Gemini 3.8 Flash. Flash is Google’s fast, cheap tier — a $19.99 Google AI Pro subscriber gets the Pro-class model. A bundle can be on the current generation and still be giving you a materially smaller model than the plan you cancelled.
ChatHub’s model page is current while its own pricing page advertises a generation older and its docs two generations older. Monica’s store listing, homepage and help centre name three different model sets. Whichever page you happened to land on decided what you thought you were buying.
Sider shipped five dated releases in 2026. TypingMind adds models within days of release. Merlin’s public changelog has a single 2026 entry. A tool that has stopped announcing is a tool whose model list will keep drifting.
FAQ
ChatHub, Sider and MultipleChat all advertise current-generation models as of 13 September 2026, and TypingMind reaches whatever your own API key does. Merlin is the outlier: its public site names nothing newer than o3, GPT 4.1, Claude 3.7 Sonnet and Gemini 2.5.
Open the model picker inside the product and read the exact names, including the size suffix. That single check answers more than any feature list, because vendors advertise model families rather than variants — “Gemini” covers both a Flash-class and a Pro-class model at very different quality and cost.
Bundles reach models through commercial APIs, so they get a model when the provider exposes it and when the bundle pays for that tier. That is usually fast for chat models and slow for anything native-only. It also means a bundle can silently swap a model out — MultipleChat's terms reserve the right to change versions, including downgrades, without notice.
Providers ship several sizes of the same model family. Google's Flash tier is optimised for speed and cost; the Pro tier is the larger, stronger model. A bundle advertising “Gemini 3.8” may be giving you either. If a page does not state the variant, assume the cheaper one and verify.
It carries the date it was last run, and the date is the point. An undated model list in this category is worthless within weeks — during 2026 alone Sider replaced its plan lineup, Monica renamed its tiers and ChatHub discontinued lifetime licences.