ChatGPT Plus at $20/Month: What You Still Get After o3-mini Went Free
16 months after o3-mini launched free, we tested whether ChatGPT Plus is still worth $20/mo. Here's what changed, and what didn't.
In January 2025, OpenAI released o3-mini to everyone (free users included). The move felt like a gut-punch to anyone paying $20 a month for access to OpenAI’s latest reasoning model. Sixteen months later, the question hasn’t gone away: what exactly are you paying for now?
We ran ChatGPT Plus through our standard value audit, comparing what free users get against what Plus actually delivers in May 2026. The answer is messier than “yes, worth it” or “no, skip it.” It depends entirely on which of your tools actually needs the upgrade.
The Free Tier Gained What Plus Lost
When o3-mini became available to free users, OpenAI didn’t leave free users with gimped access. Free tier o3-mini runs at “medium” reasoning effort, which handles the majority of real-world reasoning tasks: math, logic debugging, code walkthroughs, without the performance penalty of high effort.
Free users also got GPT-4o as the default model, web search via Bing, file uploads, and limited DALL-E 3 access. The cap is real: ~15–40 GPT-4o messages per 3-hour window. But the models themselves are no longer locked behind a paywall.
That shift forced Plus to stop being “access to the latest model” and become something narrower: higher message caps on the models everyone can use, plus a handful of features free tier doesn’t touch.
What Plus Actually Gets You (Message Caps Edition)
According to TechCrunch’s pricing breakdown, ChatGPT Plus gives you 80 messages to GPT-4o every three hours, and unlimited access to GPT-4o mini. That’s it for the headline feature. No, you don’t get o3-mini-high (the faster reasoning model). You get medium effort, same as free users, plus daily caps on how often you can trigger reasoning.
The gap between “15–40 free messages” and “80 Plus messages” sounds bigger than it feels in practice. For day-to-day writing, coding, research, and analysis, 40 GPT-4o messages often covers your morning and half your afternoon. If you’re a light user, free is probably fine.
If you’re a power user hitting caps daily, the upgrade is worth the $20. You get four times the headroom. But that’s not a feature. That’s a quantity bump.
The Real Plus Feature Set (Everything Else)
Here’s where Plus starts earning its monthly fee beyond message counts:
Advanced Voice mode with video and screen-sharing. Free tier has voice, but no video input. If you’re debugging with someone over Zoom or need to upload a screenshot for instant analysis without describing it, Plus closes a real gap.
Sora access for video generation. Free users can’t generate video. If you’re testing marketing assets or need quick visuals, this is exclusive to paid tiers.
Deep Research agent with 120 monthly queries for Plus users. This is a multi-step research tool that pulls from multiple sources, synthesizes findings, and structures output. Free users get single-turn web search. Plus gets the agentic layer.
Project workspaces with persistence and file organization. Not earth-shaking, but if you’re running parallel investigations or keeping archived research, the structure matters.
Free tier gives you raw model access. Plus gives you tooling, leverage, and headroom. Those are different things.
Who Should Upgrade (And Who Shouldn’t)
Skip Plus if you’re a casual user. Three or four queries a day, mostly writing and light coding. Free tier’s 15–40 GPT-4o messages cover you. o3-mini’s available, web search is instant, DALL-E works for quick mockups. You’re not hitting caps.
Upgrade if you’re a power user or coder. You’re running 50+ queries a day, you hit free-tier caps by lunch, and message count is the bottleneck. 80 per 3 hours takes that pressure off. The research agent and Advanced Voice are bonuses, not the reason: message count is.
Consider Claude Pro or Perplexity Pro as alternatives if you’re torn. Claude 3.5 Sonnet and Perplexity’s latest reasoning tiers offer different tradeoffs on cost, model quality, and feature set. We tested all three in our tier-comparison deep dive. Worth a 30-minute audit of your actual usage pattern before committing to Plus.
Buy Pro ($200/mo) only if you need o3-mini-high or unlimited everything. Pro is the unrestricted tier for researchers, teams, and people for whom token limits are genuinely painful. For most solo power users, Plus is the sweet spot if anything is.
What We’d Jump To If Plus Started Slipping
If OpenAI continues to push features to Pro and leave Plus with only message caps, we’d start testing Claude Pro more seriously and building multimodal workflows that don’t lock into one API. The value-per-dollar math breaks at some point, and we’re watching for it. Plus lost its “access to the latest model” moat the day o3-mini went free, so it’s competing on raw performance and message caps. If those gaps close, the decision tree changes fast.
The honest verdict: ChatGPT Plus at $20 is worth it if you hit message caps or need Advanced Voice and Sora. If you don’t, the free tier has grown teeth and it shows no signs of stopping.
Sources:
What we don't know is documented at the end of this article. We update when we learn more.