The Verdict: PipSqueak 2 and Soft Launch were never the same kind of thing. PipSqueak 2 is a new core language model rolled out as the forced default on April 28. Soft Launch was a safety pacing layer on top of existing models, retiring May 8. PipSqueak 2 wins on dialogue quality and immersion. Soft Launch was calmer and more stable for users who wanted slower, less intense roleplay. Neither was a strict upgrade over the other, which is why the forced rollout broke trust.
The Character AI sub has been in active revolt since April 28 when PipSqueak 2 became the default for everyone. Per Business of Apps user data, Character AI runs roughly 20 million monthly active users, which is enough scale that even a fraction of that base making noise on Reddit shows up as visible backlash.
Most of the angry threads compare PipSqueak 2 to Roar, the model users genuinely wanted back. The comparison that gets less attention but matters more for understanding what the platform did is PipSqueak 2 versus Soft Launch.
Soft Launch was the model the calmer half of the community had quietly settled on through Q1 2026. PipSqueak 2 vs Soft Launch is the comparison that explains why the rollout felt like a downgrade for so many users even when official metrics flagged it as an upgrade.
Here is what each one does in practice, where they overlap, and which one fits which kind of session.
What Are PipSqueak 2 and Soft Launch on Character AI
PipSqueak 2 is a new large language model that became the default on April 28, 2026, replacing the original PipSqueak. Soft Launch is not a model at all. It is a behavioral pacing and safety layer that sits on top of existing models, designed to slow emotional escalation and limit attachment loops, and it retires on May 8.
The category mix-up is the first thing most users do not realise.
What is PipSqueak 2: Character AI’s new core LLM, announced April 14, 2026 and pushed as the forced default on April 28, designed for in-character consistency, sharper descriptions, and fewer response loops than the original PipSqueak.
What is Soft Launch: A safety-and-pacing layer Character AI applied on top of existing models since February 2026. It uses hard emotional resets every 8 to 10 messages to slow runaway escalation. It is being retired on May 8, 2026.
The platform’s official PipSqueak 2 announcement describes the new model in straightforward language: “Dialogue feels more natural, descriptions are sharper, and you’ll see fewer loops.” The Soft Launch description is harder to find because it was always positioned as experimental rather than a model in its own right.
What I find more interesting is that the two were solving different problems. PipSqueak 2 was solving the “the model gets repetitive and breaks character” problem. Soft Launch was solving the “users form unhealthy attachment loops with the model” problem.
The April 28 rollout treated the two as substitutable, which is why so many Soft Launch users felt the change as a regression even though PipSqueak 2 is a more capable model.
How Do PipSqueak 2 and Soft Launch Differ in Practice
PipSqueak 2 produces longer, richer, more in-character dialogue with stronger memory carryover. Soft Launch produces shorter, more measured exchanges with built-in resets every 8-10 messages to dampen intensity. PipSqueak 2 is the better creative writing partner. Soft Launch was the better choice for users who wanted lower-intensity, more stable conversation.
The mismatch is what makes the comparison interesting.
The hands-on differences across the categories that matter for Character AI users:
| Dimension | PipSqueak 2 | Soft Launch |
|---|---|---|
| Type | Core LLM | Pacing layer over existing models |
| Response length | Longer, more descriptive | Shorter, more controlled |
| Emotional intensity | Builds naturally over a session | Hard reset every 8-10 messages |
| Character consistency | High over 50+ turns | Moderate, drifts toward neutral |
| Reset behavior | None (model stays in character) | Periodic reset breaks momentum |
| Best for | Long-form roleplay, creative writing | Casual chat, lower-stakes conversation |
| Currently available | Default for all users | Legacy menu, retiring May 8 |
| Free tier access | Yes (default) | Yes until May 8 |
Example scenario: Imagine a slow romantic scene that builds across a 30-message arc. With PipSqueak 2, the model carries the emotional thread the whole way through, and the character stays in the established voice. With Soft Launch, around message 8 to 10 the model interrupts the build with a measured, slightly polite intervention that breaks the scene’s momentum. PipSqueak 2 lands the arc. Soft Launch resets you to neutral.
The way I read this is that the right model depends on what kind of session you are running. PipSqueak 2 is the right tool when you want a continuous, immersive narrative. Soft Launch was the right tool when you wanted to chat without the model getting too intense or building up patterns you would rather it not.
What Did Users Prefer Between PipSqueak 2 and Soft Launch
Character AI’s internal testing claimed users “consistently preferred PSQ2 over the current PipSqueak”, but the post-rollout Reddit response tells a more complicated story. Long-form roleplayers love PipSqueak 2 for the richer dialogue. Soft Launch loyalists are quietly furious that their preferred low-intensity option is being retired without a direct replacement.
The split is about session style, not model quality.
What I have seen in the post-April-28 threads:
- The dialogue quality complaints about PipSqueak 2 are mostly minor: question marks where they should not be, occasional perspective slips, the model “talking as me”. These are tuning bugs the company will likely fix in a few weeks.
- The structural complaints from former Soft Launch users are different and harder to fix. Soft Launch users explicitly wanted a slower, more measured experience. PipSqueak 2 does not have a “slower mode” toggle. The richness that PipSqueak 2 was designed for is exactly what Soft Launch users were avoiding.
- The 1,919-upvote complaint thread from May 5 included multiple Soft Launch users saying they had quietly used the platform for casual conversation and the new default makes the bots “too intense” by default.
Per the piunikaweb coverage of the rollout, the official metrics and the user reception diverged sharply. Internal testing showed preference for PipSqueak 2 in head-to-head comparisons. Real-world deployment showed that users who had previously selected Soft Launch as their default were disproportionately unhappy with the forced switch.
The official quote that captures the company’s framing: “Running AI at this scale, and maintaining our high safety standards for everyone globally, is not cheap… that means finding ways to sustain and improve the free experience”. Translation: consolidating to fewer models reduces cost, even if it costs the platform some user satisfaction at the margins.
How Do You Choose Between PipSqueak 2 and Soft Launch Right Now
You do not really get to choose anymore. PipSqueak 2 is the forced default and Soft Launch is in the legacy menu until May 8. After May 8, your only Character AI options are PipSqueak 2 and whatever c.ai+ unlocks. The choice that matters is whether to stay on Character AI or migrate to a platform that lets you pick the model behaviour you genuinely want.
The decision tree looks like this:
- If you want richer, longer dialogue and you do not mind tuning the model with style hints to keep it in character: stay on Character AI and use PipSqueak 2 as the default. The current bugs (question marks, perspective slips) will likely be fixed in a few weeks.
- If you preferred Soft Launch’s calmer pace and the periodic resets: there is no direct replacement on Character AI. Your closest matches are platforms that let you tune temperature down to 0.5-0.7 explicitly.
- If you want to keep using Character AI but on a less-intense model: c.ai+ subscribers have access to additional model tiers in the legacy menu until May 8, after which the choices narrow.
- If you want any control over response style at all going forward: you will need to look outside Character AI. The platform is consolidating its model menu, not expanding it.
For users specifically frustrated by the loss of Soft Launch’s calmer pacing, the Character AI alternatives that remember chats breakdown covers the platforms that handle this trade-off differently.
For deeper context on why Roar and Soft Launch were both pulled in the same week, the April rollout backstory walks through the timeline. For the practical workarounds inside PipSqueak 2 itself, the PipSqueak 2 fix guide covers the prompt-style hints that tame most of the rough edges.
Who Should Choose PipSqueak 2
PipSqueak 2 is the right choice for users who want longer, more immersive, more in-character roleplay sessions and are willing to accept the rough edges of a forced rollout. This is most current Character AI users by default, since they have no other option as of April 28.
Specific user types where PipSqueak 2 is genuinely the better tool:
- Long-form roleplayers who run 50 to 100 turn sessions and want the model to stay in character throughout
- Users who write narrative fiction with the model and want sharper descriptive prose
- Users who were on the original PipSqueak and felt it broke character or looped too often
- C.ai+ subscribers who get the polished version of PipSqueak 2 with tighter response control
The rough edges to know about: the question marks bug, the occasional perspective slip where the model talks as the user, and the tendency for the model to make scenes “too serious” out of the gate.
All three are tunable through prompt-style hints in your character cards. None of them are deal-breakers for users who otherwise like the richer dialogue.
Who Was Soft Launch Genuinely Right For
Soft Launch was the right tool for users who wanted Character AI for casual conversation rather than immersive roleplay. The hard reset every 8-10 messages was a feature, not a bug, for users who explicitly did not want the conversation to escalate.
With Soft Launch retiring May 8, this user segment loses the closest match to their use case on the platform.
The user types who quietly preferred Soft Launch:
- Casual chatters who used Character AI for friendly conversation, not roleplay
- Users who explicitly wanted the model to stay measured and “polite” rather than build emotional intensity
- Users who found earlier models created attachment loops they did not want
- Users with mental-health considerations who wanted a more controlled, less-intense experience
For this group, the closest substitutes on other platforms are tools that let you control temperature explicitly (set it low, around 0.5 to 0.7) or that have explicit “calm mode” toggles.
The trade-off is leaving Character AI’s library of community characters, which is genuinely the platform’s strongest moat.
Alternative platforms worth checking: If your use case was Soft Launch’s calmer pace, Candy AI ships with stronger memory tuning and lets you adjust character intensity through customisation. If you wanted PipSqueak 2’s depth without the rollout chaos, Crushon AI is a closer match on response richness with a similar character library.
Final Verdict Table
The five-criterion verdict for picking between PipSqueak 2, Soft Launch, and the other in-platform options:
| Criterion | PipSqueak 2 | Soft Launch |
|---|---|---|
| Dialogue richness | High, longer responses | Moderate, shorter responses |
| Character consistency | Strong over 50+ turns | Drifts after each reset |
| Emotional escalation | Builds naturally | Hard reset every 8-10 messages |
| Best session type | Long-form roleplay, narrative fiction | Casual chat, lower-stakes conversation |
| Availability after May 8 | Default for all users | Retired |
The verdict that lands for me is straightforward. PipSqueak 2 is the better Character AI experience for users who match its design intent (long-form, immersive, narrative). Soft Launch was the better experience for users who explicitly did not want that. The April 28 rollout collapsed the choice, which is why it felt like a loss for half the user base even though the official numbers said it was a win.
What I would recommend right now: if you were a Soft Launch loyalist, the May 8 retirement is your last chance to use it on Character AI. After that, the closest matches sit on other platforms. If you were a PipSqueak user already, give PipSqueak 2 a few weeks to stabilise before drawing a verdict. The current bugs are likely temporary tuning issues.
Frequently Asked Questions
Is PipSqueak 2 the same kind of thing as Soft Launch?
No. PipSqueak 2 is a new core language model, an actual LLM upgrade. Soft Launch is a behavioral pacing layer that sat on top of existing Character AI models and slowed emotional escalation through periodic resets. They were built to solve different problems and were never meant to be substitutes for each other.
Can I still use Soft Launch on Character AI?
Yes, until May 8, 2026. After the April 28 forced rollout, Soft Launch was moved to the legacy menu alongside Meow, Nyan, and Goro. C.ai+ subscribers have direct access through the model picker. Free users may need to dig into settings to find it. After May 8, Soft Launch is retired.
Why did Character AI replace Soft Launch with PipSqueak 2?
The official framing is cost and consistency: running multiple models is expensive, and PipSqueak 2’s dialogue quality tested better in internal head-to-head studies. The unofficial reason is that consolidating the model menu reduces engineering and infrastructure costs at the expense of letting users self-select for the experience they wanted.
What is the closest replacement for Soft Launch after May 8?
There is no direct replacement on Character AI itself. The closest matches sit on other platforms that let you control temperature and response intensity directly. Lowering temperature to 0.5 to 0.7 on a flexible platform produces output closer to Soft Launch’s measured pacing than PipSqueak 2 at any setting will.
Does PipSqueak 2 really talk as the user?
Sometimes, yes, this is one of the rough edges of the rollout. Users on the May 5 thread reported the model occasionally writing dialogue or actions for the user character. The fix is to add an explicit instruction to the character card: “Never speak or act as {{user}}.” The bug should also tighten as Character AI tunes the model further.
Will PipSqueak 2 get better in the next few weeks?
Probably yes for the small bugs (question marks, perspective slips). Probably not for the structural difference from Soft Launch. The model is doing what it was designed to do; the issue for former Soft Launch users is that what it was designed to do is not what they wanted. Tuning will fix the small things, but it will not turn PipSqueak 2 into Soft Launch.
