ChatGPT Pricing in 2026: What You Actually Get on Each Tier

This page covers each tier honestly, names the model behind each one as best as can be determined from current documentation, surfaces the hidden trades that pricing pages do not advertise, and gives you the math on when each tier is worth its monthly price. API pricing follows the consumer breakdown. The closing FAQ covers the questions people search most often.

A note on data freshness: ChatGPT pricing changes more often than any other AI product on the market. OpenAI raised the Pro tier from $200 to $200 plus $100 in April 2026, retired four models from the consumer UI in February 2026, and is rolling out GPT-5.5 across tiers as this is written. The dates and figures below are accurate as of May 7, 2026. Check openai.com/business/chatgpt-pricing and platform.openai.com/docs/pricing for the live source. The maintenance schedule for this page is monthly with a weekly sweep during OpenAI announcement windows.

See also: ChatGPT 2026 overview →

The Question Pricing Pages Avoid: Which Model Are You Actually Using?

This is the question every ChatGPT user has asked at least once and every comparison article has skipped. As of March 2026, the ChatGPT model picker shows three labels – Instant, Thinking, Pro – and the actual underlying model gets selected automatically. To verify which model handled a query, you have to navigate to a Configure setting most users never open. API users always receive the specific model ID in response metadata. ChatGPT users on default settings do not.

Why this matters for pricing: a Plus subscriber pays $20 per month expecting access to GPT-5.5 because that is the headline model. The Auto selector may route their query to GPT-5.5, GPT-5.4, or a smaller fast-mode variant based on routing logic OpenAI has not fully documented. On a hard prompt, you get the flagship. On a routine prompt, you may not. The price is fixed. The value per query is not.

The table below maps tiers to underlying models as of May 2026. It is the single most-requested clarity item in the “what model am I using” search cluster, and it changes monthly.

The single most consequential editorial fact buried in this matrix: the Business tier OpenAI sells at $25 to $30 per user per month references GPT-5.2 as the underlying model on the official pricing page, even though GPT-5.5 has been the consumer flagship since April 23, 2026. Business tier customers paying for an enterprise SKU may be running on a model two generations behind the consumer Plus tier. The pricing page may not be updated in sync with model rollouts. Treat this row as volatile.

See also: AI catch ratio data →

Free Tier – The Real Cost of $0

Free is no longer pure free. As of February 9, 2026, the US Free tier displays advertisements – the first time OpenAI has placed ads in ChatGPT. The tier runs on GPT-5.3 Instant by default with GPT-5.5 Instant rolling out, capped at approximately 10 messages per 5-hour window on GPT-5.3, 3 file uploads per day, GPT Store browsing, and access to other people’s Custom GPTs.

What Free does not include: Deep Research, Advanced Voice Mode, ChatGPT Agent mode, image generation in confirmed terms, Sora video generation (which has been discontinued anyway), Custom GPT creation. Free is a try-before-you-buy surface, not a sustained workflow.

The real cost calculation for Free: if you make any commercial decision based on a ChatGPT response, the catch ratio matters. Per the Suprmind Multi-Model Divergence Index (April 2026 Edition, n=1,324 production turns), ChatGPT’s catch ratio is 0.38 – it gets caught by other models 295 times for every 111 corrections it makes. On Free without web search enabled, the citation hallucination rate is 67% per the Columbia Journalism Review audit cited in Suprmind’s AI Hallucination Rates and Benchmarks reference. Free is fine for casual queries. It is not fine for citation-dependent work, even when the price tag says zero.

Go at $8 – The Ad-Free Question

Go launched globally on January 16, 2026 after an August 2025 India-only debut. It runs on GPT-5.2 Instant, provides roughly 10x Free limits across messages and file uploads and image creation, and includes expanded memory. The catch: Go also displays ads. Paying $8 per month does not buy out the advertisement layer.

Go’s pricing is interesting only in comparison. At $8 per month, you get 10x Free capacity but still get ads. At $20 per month, Plus removes ads, upgrades you to GPT-5.5 routing, adds Deep Research, Advanced Voice, and Custom GPT creation. The math: $12 per month is the price of dropping ads and unlocking the actual feature set. For occasional users who hit Free limits and do not need Plus features, Go closes the gap. For anyone using ChatGPT for work, Go is a smaller savings than it looks.

Plus at $20 – The Tier Most People Should Pick

Plus is the entry point for serious use. It includes GPT-5.5 Instant and GPT-5.5 Thinking via the Auto selector, GPT-5.4 Pro and o3 in Flexible mode, 80 file uploads per 3-hour rolling window, 25 files per Project, 10 Deep Research queries per month, Advanced Voice Mode, image generation, ChatGPT Agent mode, Canvas, Tasks, and Custom GPT creation. No ads.

Annual billing for Plus is reported at $198 per year by independent sources – a 17.5% discount versus monthly. OpenAI does not publish the annual figure on its public pricing pages as of dossier date. Treat the $198 number as volatile until confirmed at openai.com/business/chatgpt-pricing.

The 10 Deep Research queries per month is the constraining number for research-heavy workflows. Each session takes 5 to 30 minutes and reads dozens of pages. Ten per month is two to three serious research deliverables, max. If your work demands more, the Pro tiers above are the answer.

See also: ChatGPT Deep Research deep dive →

The Two Pro Tiers – $100 vs $200

OpenAI introduced a $100 per month Pro tier on April 9, 2026, sitting between Plus and the previous Pro $200. Both Pro tiers include GPT-5.5 Pro access and the same core feature set. The differences are quantitative, not qualitative.

Pro $100 provides 5x Plus message limits, 5x Plus Codex usage, GPT-5.5 Pro access, and the standard Pro feature set. As a launch promotion, Pro $100 subscribers got 10x Codex usage through May 31, 2026 – a forcing function for sign-ups during the launch window. After May 31, the multiplier drops back to 5x.

Pro $200 provides 20x Plus message limits, 1M-token context for long-document work, the highest peak-demand priority allocation, and (until April 26, 2026) 1080p non-watermarked Sora video output up to 25 seconds. Sora’s discontinuation removes one of the Pro $200’s headline features. What remains is rate-limit headroom and context window expansion.

The math for picking between them: if your bottleneck is the number of messages per day or hour, Pro $200 buys 4x more headroom for 2x the price. If you are mostly doing Codex work or running heavy thinking sessions during the launch promo, Pro $100 buys 5x Plus capacity at half the cost of Pro $200. After the May 31, 2026 promo ends, the Pro $100 value compresses. For users hitting Plus limits frequently but not constantly, Pro $100 is the natural upgrade. For users who hit ceilings every day, Pro $200 is still the right tier.

Business and Enterprise – What You Are Actually Buying

Business (formerly ChatGPT Team, renamed August 2025) is $30 per user per month billed monthly or $25 per user per month billed annually. The annual price is the only OpenAI tier where the discount is published openly. Business adds shared workspaces, SAML SSO, no model training on your data, SOC 2 Type 2 compliance, the Codex agent, Deep Research, 32K context for non-reasoning models, and 196K context for reasoning models. As of dossier date, Business does not include SCIM provisioning or ISO 27001 certifications – those are Enterprise.

Business is the right tier for teams under 50 people who need shared workspaces and basic security but do not need provisioning automation. The 32K non-reasoning context is the constraining limit. For teams routinely working with documents larger than 24,000 words or codebases larger than a few hundred files, the context window will become the bottleneck before the per-user price does.

Enterprise is custom-priced. Independent estimates land in the $40 to $60 per user per month range, but OpenAI does not disclose. Enterprise adds ISO 27001/27017/27018/27701 certifications, SCIM provisioning, enterprise key management, role-based access control, an analytics dashboard, IP allowlisting, data residency options across the US, EU, UK, JP, CA, KR, SG, IN, AU, and UAE, a global admin console, 24/7 priority support, and custom legal terms.

For procurement teams: Enterprise is also the only tier where you can negotiate around model routing transparency, data retention duration, and custom SLAs. If your organization has a regulated compliance posture or data residency requirement, the negotiated terms matter more than the per-seat math.

API Pricing – The Developer Alternative

The API runs on the same GPT models that power ChatGPT, but bills per token instead of per seat. For workloads with predictable cost-per-query, the API can be substantially cheaper than ChatGPT Plus. For workloads with high per-query token volume, it can be substantially more expensive.

Source: openai.com/api/pricing as of 2026-05-07.

The API also offers three modifier tiers. Batch processing runs at 50% of standard rates with 24-hour async processing. Flex offers lower cost and slower response with occasional resource unavailability. Priority is pay-as-you-go at 2.5x standard rates for guaranteed throughput.

The 33x Cost Ratio Most Pages Skip

GPT-4o mini at $0.15 per 1M input tokens is roughly 33 times cheaper than GPT-5.5 at $5.00 per 1M input. The output ratio is similar – $0.60 versus $30.00, which is 50x. For workloads that do not require flagship capability, the older multimodal model is still the cost-efficient default.

When does this matter? Specifically: high-volume routine work like classification, extraction from structured data, content moderation, simple summarization, and any task where you can verify quality with rules rather than trust. The hallucination profile matters less when the task is constrained. GPT-4o mini’s Vectara old-dataset hallucination rate of 1.7% is lower than GPT-5’s >10% on the harder Vectara new dataset. For specific workloads, the smaller cheaper model is also the safer choice.

The case for paying 33x more: open-domain reasoning, multi-step planning, agentic computer use (where GPT-5.4 introduced the capability), long-context recall, and any work where the question itself can change shape during the conversation. Flagship-level capability earns the price. Routine work does not.

Cached Input – The Pricing Lever Most API Users Ignore

Cached inputs cost 90% less than fresh inputs across the GPT-5.x family. GPT-5.5 cached input is $0.50 per 1M versus $5.00 fresh. GPT-5.4 cached is $0.25 versus $2.50. GPT-4.1 cached is $0.50 versus $2.00.

Caching applies when a request reuses a substantial portion of a recent prior request’s prompt material. The OpenAI API automatically caches frequently used prompt prefixes. For workflows where you prepend a long system prompt, large reference document, or stable instruction block to every request, cached input pricing is a 10x cost reduction with no engineering work.

Workflows that benefit most: agentic loops with consistent system prompts, RAG pipelines reusing the same document context, customer-support bots with stable persona instructions, any production workload with templated prompts. If your API spend is dominated by input tokens and your prompts include stable prefixes, caching is the largest single lever for cost control.

Recent Pricing Changes Timeline

The pace is the relevant fact. Six material pricing or model-availability changes in 18 months. Three in the last 90 days. Any pricing comparison built today will need to be re-verified before any decision over $1,000 per month.

Geographic Restrictions and the Italy Precedent

ChatGPT is available in most countries. Sanctioned jurisdictions (Iran, North Korea, Cuba, in some cases Russia) are blocked per US export control compliance. Mainland China is unavailable. UK, EU, and most of Asia and Latin America have access.

The Italian DPA temporarily banned ChatGPT in March 2023 over GDPR concerns. OpenAI complied within the deadline by introducing GDPR-specific privacy disclosures, age verification, and a training opt-out tool. Service was restored in May 2023. The action did not result in a formal GDPR fine. The episode established that EU data protection authorities can act against AI systems without waiting for EU AI Act enforcement, which is a procurement risk to factor into European deployments at scale.

For Enterprise customers, OpenAI offers data residency in the US, EU, UK, JP, CA, KR, SG, IN, AU, and UAE. If you have a regulatory obligation to keep data in a specific jurisdiction, this is the tier where that becomes possible to negotiate.

FAQ

Frequently Asked Questions

Stop guessing. Start cross-checking.

Suprmind runs your prompt across ChatGPT, Claude, Gemini, Grok, and Perplexity in parallel. See where they agree, where they disagree, and which insights only one model surfaced — before you act.