Aikeedo v4.1 Is Live: Thinking Controls, Session Billing, Private Chats, and a Full Model Refresh

v4.0 was our biggest release ever. v4.1 is the one that sharpens it.

This update is less about a single headline feature and more about giving both you and your users more control — over how models think, how usage is billed, how privacy is handled, and how secure the whole platform is underneath. Plus the largest model refresh we have ever shipped in a single point release.

Aikeedo v4.1 is live.

Here is everything that is new.

Reasoning Controls: Let Users Decide How Hard the Model Thinks

Modern models can think before they answer — but until now, that was an all-or-nothing setting an admin flipped in the background. v4.1 turns reasoning into a real, user-facing control right inside the chat.

Your users now get a Claude-style model selector with two new controls:

• A thinking toggle — turn extended reasoning on or off per message
• Effort tiers — choose how much effort the model puts in: low, medium, high, or max

These settings are applied per message, so a user can fire off a quick question with thinking off, then switch to high effort for a hard problem in the very next message. Under the hood, Aikeedo maps each choice to the correct provider-specific parameters across OpenAI, Anthropic, xAI, Cohere, and custom models — so the same simple toggle works no matter which model your user picked.

For operators offering bring-your-own-key or custom models, you can configure the full reasoning profile per model from the admin panel. Billing is unchanged — usage is still metered from the actual API response, so nothing about your cost accounting shifts.

Privacy: Temporary Chats and Account Deletion

Two features in v4.1 are aimed squarely at giving your users control over their own data — which is increasingly what they expect, and increasingly what regulations require.

Temporary Chats (Incognito mode). Users can now start a temporary chat that is not saved to their history. It is the same pattern people already know from consumer AI tools: a private session for one-off questions, sensitive drafts, or anything they simply do not want cluttering their conversation list. When the chat is closed, it is gone.

Account deletion. v4.1 adds a complete self-service account deletion flow. Users get a clear entry point in their account settings, a confirmation step to prevent accidents, and proper cleanup of associated data — subscriptions, content, and the rest — when they delete. For operators running platforms in regions with strict data-protection rules, this closes a real compliance gap without you having to build it yourself or handle deletions manually.

Session-Based Usage Mode: A "Limitless Feel" for Your Users

One of the biggest additions in v4.1 is a new, site-wide usage mode that changes how consumption is presented and paced on your platform. You choose the mode that fits your business:

• Credit metering (unpaced) — the default, and exactly how Aikeedo works today: visible credit balances drawn from a monthly pool.
• Session-based usage — a new mode that hides raw credit balances, displays usage in your own currency, and paces consumption across fixed session windows instead of a single monthly pool.

Session mode is built for operators who want their platform to feel less like a metered utility and more like an always-available product. Instead of watching a credit counter tick down, users see a currency-based usage view and a session window that resets on a schedule you configure (anywhere from 1 to 24 hours). When a user hits a session cap, they get a clear "resets in X hours Y minutes" message rather than a hard wall.

Credits remain the internal unit everywhere, so your cost accounting, add-ons, and custom credit purchases keep working exactly as before — this is purely about how usage is presented and paced to the end user. The enforcement runs through a central usage gate that also respects bring-your-own-key setups (zero-cost OpenAI and Anthropic calls skip the gate) and works across both the core app and the Chatbots plugin.

To support all of this, the chat box itself now surfaces the user's current usage limit inline, so people always know where they stand before they hit a wall.

The Biggest Model Refresh Yet

v4.1 brings the model catalog fully up to date across text, image, and video — so your users get access to the newest models the moment they matter.

New text models:

• OpenAI GPT 5.5 and GPT 5.5 Pro
• Anthropic Opus 4.8 and Opus 4.7
• Anthropic Sonnet 5
• Anthropic Claude Fable 5
• xAI Grok 4.3

New image and video models:

• GPT Image 2 — OpenAI's latest image model
• Seedream 5.0 Lite — fast, lightweight image generation
• Seedance 2 — next-generation video
• Kling 3.0 — high-quality video generation

To keep the catalog clean and current, v4.1 also retires a batch of superseded models. Removed: Grok 4.1 Fast, Grok 4 Fast, Grok 4, Grok 3, Grok Code Fast, Grok Imagine Image Pro, Anthropic Opus 4.0 and 4.1, and OpenAI GPT-5, GPT-5 Pro, GPT-5 Mini, and GPT-5 Nano. Sora 2 and Sora 2 Pro are now deprecated. Combined with the redesigned Model Registry from v4.0, you can keep activating new models the day they drop — without touching source code.

Behind the scenes, v4.1 also lays new video reference infrastructure for fal.ai video models, adding support for reference frames and image/element tags in prompts — groundwork that makes richer, guided video generation possible in the tools built on top of it.

More Features Worth Knowing About

v4.1 also ships a set of improvements that clean up real friction operators and users have run into:

Configurable "Detailed Pricing Table" link. The pricing button on the in-app Plans page used to point at the built-in landing page anchor — which broke when operators replaced the landing page with an external marketing site like WordPress. You can now set a custom URL for that button from the admin panel, so a conversion-critical link never dead-ends again.

Mobile chat image upload. Fixed image upload from the photo library on mobile, so uploading a picture straight from a phone works as expected.

Mercado Pago sandbox fix. Corrected a bug where the Mercado Pago gateway could get stuck in sandbox mode, blocking real payments for operators in Latin America.

Migration and import fixes. A round of fixes to the migration plugin, including resolving cases where ChatGPT conversations were not being imported during OpenAI export migrations — so users bringing their history over from other tools get a complete migration.

Navigation collapse persistence. Collapsed sidebar sections now stay collapsed between visits, a small but welcome quality-of-life fix.

Plugin installation fixes. Improvements to admin ZIP upload so installing plugins from the admin panel is more reliable.

Under the Hood: A Security and Reliability Pass

A large part of v4.1 went into hardening the platform. If you run Aikeedo in production, these are the changes worth knowing about.

Security hardening:

• Secure auth cookies — authentication cookies are now set with HttpOnly and SameSite flags
• JWT expiry on all auth tokens, including "remember me" sessions, so tokens no longer live indefinitely
• JWT from query strings is now gated, closing a token-leakage vector
• Workspace isolation fix in the user middleware, tightening multi-tenant separation
• Password minimum-length validation on account creation and changes
• Admin upload validation — file extensions are now checked against a whitelist on admin uploads, closing a gap where upload file types were not being validated

Reliability and observability:

• Fixed credit-deduction atomicity in message generation, resolving a mid-stream flush issue so credits are accounted for reliably during streaming
• Separated error logging into its own stream with rotating file handlers, plus improved production logging in the exception middleware — so real issues are easier to find and diagnose
• Streaming robustness — incomplete streaming generators in the Completions API are now handled gracefully instead of throwing fatal errors, and malformed UTF-8 no longer triggers iconv() warnings in the embedding service

Chat internals:

• Architectural improvements to the chat tooling, including correct rendering of error and source message parts in the UI, a fix for legacy chat-history tool content, and proper injection of assistant instructions for OpenAI and xAI models

Add in triage of a batch of community-reported GitHub issues — several fixed, others closed as not reproducible — and v4.1 is meaningfully more solid to run than the release before it.

How to Upgrade

If you are on an active license with updates included, you can pull v4.1 now. Full upgrade instructions are in the documentation.

If you are evaluating Aikeedo for the first time, v4.1 is a great place to start — the newest models, the new usage and reasoning controls, the privacy features, and the security work are all included out of the box.

Try the live demo to see it in action, or head to our pricing page if you are ready to get started.

More to come.