OpenAI CEO Sam Altman says “compute-intensive” ChatGPT features are rolling out in the coming weeks. Some will be limited to $200/month ChatGPT Pro subscribers, and a few may carry extra fees. The company frames this as a cost reality, not a permanent wall, and says it still aims to drive prices down over time.
Will new ChatGPT features be paywalled?
Yes. Altman says some upcoming “compute-intensive” features will initially be Pro-only, with possible extra fees on select products. The company’s stated aim is to lower costs and broaden access over time, but near-term rollouts will favor Pro users.
What exactly did Altman say and when?
On September 21, Sam Altman posted on X that OpenAI will ship “new compute-intensive offerings” over the next few weeks. Because they’re costly to run, some features will start as Pro-only, and some products will have additional fees. He added that OpenAI’s goal remains to make AI widely available by pushing costs down.
Independent reports and write-ups (BI, Tom’s Guide, PCWorld) repeated the same core points: new features, heavy compute needs, Pro-first access, and potential surcharges.
What is ChatGPT Pro today?
ChatGPT Pro launched in December 2024 at $200/month. It’s aimed at power users who hit usage caps or need the highest-end models and voice tools. OpenAI’s announcement lists unlimited access to its top models and advanced voice, plus a higher-compute “pro mode” for the o1-family. The Help Center and pricing pages now describe Pro with unlimited or extended access to GPT-5 and priority throughput.
Why that matters now: If OpenAI is about to launch compute-hungry features, Pro is the natural home for early access. The plan already targets users who need deeper reasoning, more tokens, and higher reliability during peak demand.
What new features might be paywalled?
OpenAI didn’t list them yet, but there are credible signals:
- AI browser: Reuters reported OpenAI is preparing a browser that bakes ChatGPT into the browsing experience and could include agent-like automation. That’s a prime candidate for Pro-first access.
- Agent upgrades: Operator, OpenAI’s task-performing agent, has been integrated into ChatGPT as “agent mode.” More capable agents typically burn more compute, which fits a Pro-tier rollout.
- “Thinking” controls: Tech press has covered new thinking-time options for GPT-5, with Pro users getting heavier modes. Expect Pro-only toggles for deeper reasoning on big tasks.
- Media generation: When OpenAI debuted Sora, higher limits and better quality were tied to paid tiers. If new video or multimodal tools expand, they may land in Pro first.
A broader business backdrop supports this: OpenAI has reportedly hit a $12B annualized revenue run rate, driven heavily by subscriptions. Prioritizing Pro for expensive features is consistent with funding the compute behind them.
Why Pro-only? The compute math, simply explained
Reasoning-heavy tasks don’t just send one request and call it a day. They spin through chains of thought, tool calls, and verification passes. That’s a lot of tokens and GPU time, which means real money. Even small latency improvements at scale need massive infrastructure. Altman’s post is essentially a heads-up: if you want the biggest leaps first, they’ll live where the compute budget is.
How this affects Free and Plus users
- Free: You’ll keep access to core chat and the flagship model with standard limits. Cutting-edge features may arrive later or with lower quotas.
- Plus ($20): Expect solid access to GPT-5 and mainstream tools, but fewer knobs for deep reasoning or heavy agent runs. Pro may unlock priority throughput and extra “thinking” or automation.
- Pro ($200): Likely first in line for new agents, browser-level features, higher caps, and “heavy” reasoning modes.
Reality check: OpenAI often pilots big features at higher tiers and then steps prices down as infrastructure scales. That pattern matches Altman’s stated intent.
ChatGPT tiers compared
| Tier | Price | Models | Voice/Video | Agents | Access & Limits |
|---|---|---|---|---|---|
| Free | $0 | GPT-5 (standard access) | Basic voice | Basic tools | Limited usage, standard throughput |
| Plus | $20 | GPT-5 with extended limits | Standard + video/screen-share | Access to agent | Higher limits than Free |
| Pro | $200 | GPT-5 with priority + heavy modes | Advanced voice, higher media limits | Agent with more headroom | Priority traffic, early feature access |
Competitive context: Google and others
Google is pushing richer AI in consumer products and coding tools, with some features reserved for paid tiers. If OpenAI’s browser lands and agents keep improving, Pro-first access helps OpenAI hold ground with power users who might otherwise drift to rival ecosystems.
Should you upgrade? A simple decision checklist
Consider Pro if you:
- Regularly hit rate limits or long wait times
- Run complex coding, data, or research tasks that benefit from deeper “thinking”
- Need agentic workflows that complete multi-step jobs
- Plan to test early features for your team or product
Stick with Plus/Free if you:
- Mostly draft, summarize, and brainstorm
- Only occasionally need heavy runs
- Don’t need priority throughput or early access
Pricing watch: where extra fees may show up
- High-compute runs: deep reasoning “heavy” modes on large inputs
- Automation: long-running agents that browse, book, or transact
- Media: high-res or longer video generations
Nothing official yet, but these are the usual compute hotspots.
FAQ
Why are compute-heavy features paywalled at first?
Running deep reasoning, agents, and media generation consumes a lot of GPU time. Paywalling early helps fund capacity and stabilize quality before broader rollout.
What exactly do Pro users get today?
Per OpenAI, Pro includes unlimited access to top models, advanced voice, priority throughput, and higher-compute modes, plus opportunities to test new features.
Is OpenAI doing this for revenue?
Revenue matters. OpenAI reportedly crossed $12B annualized revenue mid-2025, with subscriptions playing a key role. Compute-intensive features likely need Pro economics.
Will there be extra fees beyond $200?
Altman said some new products may carry additional fees. Expect this around long-running agents or media tasks that burn lots of compute.
What about GPT-5 and thinking controls?
Reports highlight new thinking-time controls, with heavier modes for Pro. This is consistent with a Pro-first approach to deep reasoning.
When will these features launch?
Altman said “over the next few weeks.” That suggests staged rollouts rather than a single big drop
Featured Answer Boxes
What did Sam Altman announce about new ChatGPT features?
He said OpenAI will launch “compute-intensive” offerings in the coming weeks. Some features will start as Pro-only, and some products could carry extra fees due to high GPU costs. The stated goal is still to lower prices over time.
How much is ChatGPT Pro and what’s included?
ChatGPT Pro costs $200/month. OpenAI lists unlimited access to top models, priority performance, advanced voice, and higher-compute modes, with early access to select features over time.
Will Free or Plus users lose features?
There’s no indication of removals. Expect new, heavy features to debut on Pro first, then broaden if costs fall. Free and Plus keep core functionality with lower limits.
Is an OpenAI browser coming?
Reuters reported OpenAI is prepping an AI-powered browser that integrates ChatGPT and agent capabilities. If launched, it’s a likely Pro-first tool.
Are agents part of this push?
Yes. OpenAI’s Operator agent is integrated into ChatGPT as “agent mode” and is likely to see rapid upgrades that consume more compute.
Source: OpenAI | X (formerly Twitter) Altman’s X post
