Somewhere inside a data center in Abilene, Texas, running on more than 100,000 NVIDIA H100 GPUs, OpenAI's next flagship AI model has been sitting in safety evaluation since March 24, 2026 — the day its pretraining officially completed. OpenAI CEO Sam Altman called it 'a very strong model that could really accelerate the economy.' President Greg Brockman described it publicly as representing 'two years of research' with a 'big model feel — it's not an incremental improvement.' The internal codename is Spud. The public name, when it ships, will either be GPT-5.5 or GPT-6, depending on how large a performance leap OpenAI's own team decides to claim. As of April 15, 2026, prediction markets on Polymarket assign 78% probability that this model releases by April 30, and over 95% probability by June 30. The 3-to-6-week safety evaluation window OpenAI has followed for every major recent model puts the earliest realistic release date between April 14 and May 5. Sources: The Information, March 24, 2026 (pretraining completion); Sam Altman, public statement to employees, March 24, 2026; Greg Brockman, Big Technology podcast, April 7, 2026; Polymarket contract data, April 14, 2026.

This creates an intensely practical problem for millions of Americans. The question isn't 'when does GPT-6 come out' — it's the human question underneath it: should I pay for an AI subscription right now, or wait two to three weeks for what might be the biggest model upgrade since GPT-5? And if I already pay, should I upgrade to OpenAI's new $100 per month Pro tier, or does that not make sense until after GPT-6 arrives? This guide exists to answer those questions concretely — not with benchmark tables, but with decision scenarios mapped to the way real people actually use AI. We cover the current AI landscape as of April 15, the confirmed and unconfirmed facts about Spud, what the new $100 Pro tier actually gives you, and a clear framework for five different types of AI users.

What Is Actually Confirmed About GPT-6 Spud (vs What Is Rumor)

The signal-to-noise ratio on GPT-6 coverage is currently terrible. Social media is full of 'CONFIRMED: GPT-6 launches April 14 with 2 million token context and 40% better performance' — most of which is either inference dressed as fact or outright fabrication. Here is a clear separation of what is confirmed versus what is speculation, so you can make an actual decision.

The most honest framing: GPT-6 is real, it is close (weeks, not months), it is not incremental, and everything beyond that is speculation. The expected window is April 14 through May 5, with Polymarket's liquidity concentrating around late April. The feature claims floating online are mostly extrapolations from competitive benchmarks — the 2M context window is competitive inference (Gemini 3.1 Pro already has 1M), not a confirmed spec. When the official blog post, model card, and API documentation appear, all of that will be settled immediately. Until then, the confirmed facts are sufficient to make a rational subscription decision.

The April 15, 2026 AI Landscape: What Each Subscription Buys You Right Now

Before evaluating whether to wait for GPT-6, you need to know what the current models actually offer. The frontier is genuinely fragmented in April 2026 — no single model dominates every category.

Sources: OpenAI pricing page, April 2026; Anthropic pricing page, April 2026; Google AI Studio pricing, April 2026; BenchLM AI leaderboard, April 8, 2026; SWE-bench Verified public leaderboard, April 2026.

OpenAI's New $100/Month Pro Plan: What Changed on April 9, 2026

OpenAI introduced a new $100 per month Pro plan on April 9, 2026 — splitting what was previously a single $200 Pro tier into two options. This is a significant pricing change that directly affects the GPT-6 upgrade decision. The new $100 tier is positioned as 'built for longer, high-intensity Codex sessions' and sits between ChatGPT Plus ($20) and the original $200 Pro tier. Understanding exactly what the two tiers include — and what they do not — is essential to evaluating whether the timing of the new tier is relevant to your decision. Source: OpenAI release notes, April 8-9, 2026.

• ChatGPT Plus ($20/month): Unlimited access to GPT-5.4 Thinking and Instant in ChatGPT, deep research, file analysis, web search, voice mode, image generation, and access to GPT-5.4 Pro in limited quantities. 5x Codex session usage compared to free. This is the right plan for the vast majority of users — it covers all general-purpose professional work without constraint. Source: OpenAI release notes, April 9, 2026.
• ChatGPT Pro New ($100/month): Everything in Plus, plus unlimited GPT-5.4 Pro access (vs limited in Plus), and 10x Codex usage versus Plus (up from 5x standard usage). For context: Codex is OpenAI's agentic coding agent that runs multi-step software engineering tasks autonomously in a cloud sandbox. The $100 tier specifically targets developers and engineering teams that are running sustained, long-horizon Codex sessions daily. Source: OpenAI release notes, April 9, 2026.
• ChatGPT Pro ($200/month — Pro 20x): The original high-compute tier remains available for maximum usage, with the Codex promotion extended through May 31. Designed for the heaviest possible professional workloads. Source: OpenAI release notes, April 9, 2026.
• The key insight about timing: The new $100 tier becomes substantially more valuable the moment GPT-6 ships, because GPT-6 Pro access will likely be gated at that tier first — following the historical pattern from every GPT-5.x release. Subscribing to the $100 tier now means you pay for current GPT-5.4 Pro capability while waiting 2-4 weeks for GPT-6. Subscribing after GPT-6 launches means you access the new model immediately at its best capability level. For most users, waiting 2-4 weeks to subscribe is the rational choice unless Codex usage is already a daily bottleneck. Source: historical OpenAI tiering patterns (GPT-5.2 Pro, GPT-5.4 Pro rollouts); Polymarket, April 14, 2026.

5 Decision Scenarios: The Honest Answer for Each Type of AI User

Scenario 1: You use AI free tiers occasionally and are curious whether to subscribe

Decision: Wait 2-4 weeks. The free tiers of ChatGPT, Claude, and Gemini are more capable in April 2026 than paid tiers were 18 months ago. If your use is truly occasional — a few queries per day, casual research, help with emails — you are not hitting meaningful limits on free access. Waiting costs you nothing. When GPT-6 launches, the Plus/Pro tiers will immediately include it, so subscribing then means you skip the current generation entirely and enter on the new baseline. The only exception: if you have a time-sensitive project in the next 2-3 weeks that requires deep research, sustained Codex sessions, or heavy file analysis, subscribe to ChatGPT Plus or Claude Pro now and get immediate access to the best current models.

Scenario 2: You use AI daily for professional work and are currently on ChatGPT Plus ($20)

Decision: Stay on Plus for now. ChatGPT Plus at $20 gives you GPT-5.4 Thinking — OpenAI's current frontier reasoning model — with no meaningful daily use limits for most professionals. It leads on knowledge work tasks (83% GDPval, matching or exceeding human professionals in 83% of comparisons across 44 occupations), computer use (75% OSWorld), and deep research quality. The $100 Pro tier does not fundamentally change the model you interact with in ChatGPT — it primarily adds more Codex session time and unrestricted GPT-5.4 Pro access. For writing, analysis, planning, and research, Plus is sufficient. Wait until GPT-6 launches to evaluate whether an upgrade makes sense based on GPT-6 Pro's actual capabilities. Sources: OpenAI GDPval benchmark, March 5, 2026; OpenAI release notes, April 9, 2026.

Scenario 3: You are a developer or engineer who uses Codex heavily

Decision: Upgrade to the new $100/month Pro tier now, but understand what you are paying for. If you are hitting Codex session limits on ChatGPT Plus and that is actively slowing your work, the $100 tier directly solves that problem with 10x the Codex usage. GPT-5.4 is already the leading model for autonomous multi-step computer use (75% OSWorld), and GPT-5.3-Codex — the underlying agentic coding model — holds state-of-the-art performance on SWE-Bench Pro and Terminal-Bench 2.0 as of April 2026. This tier will almost certainly be the access path to GPT-6's new agentic capabilities at launch, so upgrading now means uninterrupted access when GPT-6 ships. The math: $100 per month for 4 hours per day of Codex-accelerated engineering work is a small fraction of the engineering time you are recapturing. Sources: OpenAI SWE-Bench Pro results; OpenAI release notes, April 9, 2026; Terminal-Bench 2.0 leaderboard, March 2026.

Scenario 4: You are deciding between Claude Pro vs ChatGPT Plus right now

Decision: This depends on your primary use case, not on GPT-6's arrival. Claude Pro leads for coding quality (Claude Opus 4.6 at 80.8% SWE-bench Verified — OpenAI stopped self-reporting Verified scores due to contamination concerns, with independent leaderboards placing GPT-5.4 at approximately 78-80%) and long-form writing — Claude Opus 4.6 consistently wins head-to-head user preference tests on prose quality, stylistic accuracy, and following complex multi-part instructions. ChatGPT Plus leads for versatility, knowledge work breadth (GDPval 83%), computer use, and the adaptive memory system that automatically accumulates your preferences across conversations. Important update: Claude launched automatic Chat Memory for all plans — including the free tier — on March 2, 2026. Claude now synthesizes conversations every 24 hours and builds a persistent memory profile, visible and editable at Settings > Capabilities. Both platforms now offer automatic cross-conversation memory; the key difference is that ChatGPT's memory has had more time to accumulate a detailed profile if you are a long-term user. If your primary work is coding or writing, Claude Pro is the better choice regardless of GPT-6's arrival. If your work is mixed and you want one AI that handles everything, ChatGPT Plus is better. GPT-6's arrival will reset the coding and writing benchmarks, but it will not change the fact that these models serve different workflows. Sources: SWE-bench Verified public leaderboard, April 2026; OpenAI GDPval benchmark, March 2026; Anthropic Chat Memory announcement, March 2, 2026; independent user testing, LumiChats research.

Scenario 5: You are on Gemini Advanced (Google AI Pro at $19.99) and wondering whether to switch

Decision: Stay on Gemini Advanced unless your work is primarily coding or conversational writing. Gemini 3.1 Pro leads the frontier on scientific and PhD-level reasoning (94.3% GPQA Diamond — a consistent 2-3 point lead over every competitor in independent evaluations), handles native multimodal analysis (audio and video natively without transcription), and integrates directly with Google Workspace through Docs, Sheets, Slides, and Drive. If you work in Google's ecosystem, Gemini's Personal Intelligence integration — which connects to your Google account context — is a feature no ChatGPT or Claude subscription can match. GPT-6 will likely improve across all categories, but it will not automatically become a better fit for Google Workspace users, scientific researchers, or people who rely on native audio/video analysis. Gemini 3.1 Ultra (Gemini's next tier, at 2M tokens) is also expected Q2 2026 — so Google's roadmap is equally active. Sources: GPQA Diamond leaderboard, April 2026; Google AI Pro announcement, February 2026; Bloomberg Gemini-Apple deal reporting, 2026.

The Historical Pattern: What Actually Happened When Previous Models Launched

GPT-6's arrival will follow a predictable pattern based on every major model launch since GPT-5 in January 2026. Understanding this pattern removes the anxiety from the decision entirely.

• Day 0 of GPT-6 launch: ChatGPT Plus subscribers get immediate GPT-6 Thinking access. This is confirmed by OpenAI's pattern: every GPT-5.x Thinking model replaced the previous Thinking model as the default for Plus subscribers at launch, with no additional payment or upgrade required. Source: OpenAI release notes for GPT-5.2, GPT-5.3, GPT-5.4 launches.
• Day 0: API and Codex: GPT-6 will be available in the API and Codex simultaneously or within days. GPT-5.4 launched in ChatGPT, the API, and Codex on the same day (March 5, 2026). The same architecture is expected for GPT-6. Source: OpenAI release notes, March 5, 2026.
• First 2-4 weeks: GPT-6 Pro gated to $100+ tiers. Following the GPT-5.4 Pro pattern, the maximum-capability version of GPT-6 will likely be restricted to Pro tier subscribers initially. This is the specific reason developers considering the $100 tier should time their subscription to coincide with or just precede the GPT-6 launch. Source: OpenAI tiering history, GPT-5.4 Pro rollout, March 5, 2026.
• First 2-4 weeks: Free tier gets limited access. Free ChatGPT users will get GPT-6 access at reduced rate limits, following the pattern where GPT-5.4 mini became the free tier fallback. If you are currently on the free tier, GPT-6 will reach you — it just will not be unlimited. Source: OpenAI pattern from GPT-5.3 Instant Mini rollout, April 8, 2026.
• After launch: The current best models do not disappear. GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro will remain available and fully functional after GPT-6 ships. OpenAI typically retires models 3 months after replacing them as default (GPT-5.2 retiring June 5, 2026). There is no urgency to subscribe before GPT-6 to 'lock in' access to current models. Source: OpenAI retirement schedule, April 2026.

Claude and Gemini After GPT-6: Will They Still Matter?

The honest answer is yes — and understanding why matters for the subscription decision. The past 18 months have proven that AI model leadership does not stay with one company for long. When GPT-5.4 launched on March 5, 2026, it was widely described as re-establishing OpenAI's lead. Within weeks, independent benchmarks showed Gemini 3.1 Pro (released February 2026) leading on 13 of 16 major benchmarks, and Claude Opus 4.6 holding the top position on coding. GPT-5.4 leads strongly on knowledge work (GDPval 83%) and computer use (OSWorld 75%), but the competitive gaps are measured in percentage points, not generations. The realistic projection: GPT-6 will lead most benchmarks for approximately 6 to 10 weeks before Anthropic and Google respond. Anthropic's Claude Mythos, described as 'the most powerful AI model ever developed' in a data leak from March 26, 2026, is expected Q2 2026. Google's Gemini 3.1 Ultra with 2M context is also expected before mid-year. Q2 2026 is being called the most competitive quarter in AI model history by multiple analysts. Source: LumiChats reporting; Nipralo Technologies analysis, April 2026; AI researcher commentary, HumanX conference, April 2026.

The One Scenario Where Waiting for GPT-6 Is Clearly Wrong

There is one case where waiting is clearly the wrong call: if you have specific, high-value work that needs to happen in the next three weeks and you are currently using free-tier AI to do it. Free-tier AI in April 2026 hits rate limits that materially slow professional work — deep research is gated, context windows are limited, Codex is unavailable on free. If you have a project, a deadline, a code repository that needs agent-based refactoring, or a research deliverable due before May, subscribe to ChatGPT Plus ($20) or Claude Pro ($20) today. The productivity gain from current frontier models over free-tier models is substantial and immediate. Three weeks of Plus subscription costs $5. The productivity return on removing AI rate limits from your professional workflow is not close. GPT-6 will arrive shortly after, and your Plus subscription will automatically include it.

Frequently Asked Questions