Last Updated: November 1, 2026

GPT-5 is no longer OpenAI's newest model. OpenAI launched GPT-5 on August 7, 2025, shipped six GPT-5.x updates through July 2026, and then replaced the whole series in ChatGPT with GPT-6 in October 2026: GPT-6 Sol for paid plans from October 7 and GPT-6 Luna for Free and Go users from October 8, according to OpenAI's community announcement.
GPT-5 still matters for two reasons. Its August 2025 launch benchmarks remain the reference point people compare newer models against, and the original GPT-5 API models keep working until December 11, 2026, when OpenAI's deprecations page says they shut down.
This page keeps every GPT-5 statistic that can be traced to an official source, labels each one with its date, and adds what happened after: the GPT-5.x timeline, GPT-6, API availability and what GPT-5 cost compared with the models that replaced it.
Free guide: Which AI Tool Should You Actually Use? Join 2,000+ business leaders who read AI Business Weekly for free, and get the guide in your inbox right away.
Table of Contents
Is GPT-5 Still OpenAI's Latest Model?
No. GPT-6 is the current model family in ChatGPT as of November 2026. OpenAI started rolling out GPT-6 Sol to Plus, Pro, Business and Enterprise on October 7, 2026 and GPT-6 Luna to Free and Go on October 8, 2026, per the OpenAI Developer Community announcement.
The GPT-5 series ran for about 14 months, from August 2025 to October 2026. The table below lists each release with the date taken from OpenAI's own announcements or API changelog.
Model | Released | Headline fact | Source |
|---|---|---|---|
GPT-5 | August 7, 2025 | 94.6% on AIME 2025 without tools | |
GPT-5.1 | November 13, 2025 (API) | Released to the API platform | |
GPT-5.2 | December 11, 2025 | Instant, Thinking and Pro; 100% on AIME 2025 (Thinking, no tools) | |
GPT-5.3-Codex | February 5, 2026 | Coding agent for long, multi-step computer tasks | |
GPT-5.4 | March 5, 2026 | 75.0% on OSWorld-Verified computer use | |
GPT-5.5 | April 23, 2026 (ChatGPT), April 24 (API) | 1M token context window in the API | |
GPT-5.6 Sol, Terra, Luna | July 9, 2026 | Three tiers by capability and cost | |
GPT-6 Sol and Luna | September 22, 2026 (API); October 7 and 8, 2026 (ChatGPT) | Replaced GPT-5.x in ChatGPT |
What Replaced GPT-5 in ChatGPT?
GPT-6 replaced the GPT-5.x models in ChatGPT in October 2026. Paid plans (Plus, Pro, Business, Enterprise) get GPT-6 Sol, and the Free and Go plans get GPT-6 Luna, according to OpenAI's rollout notice.
ChatGPT plan | GPT-6 model | Rollout start |
|---|---|---|
Plus, Pro, Business, Enterprise | GPT-6 Sol | October 7, 2026 |
Free, Go | GPT-6 Luna | October 8, 2026 |
OpenAI's only speed figure for the switch compares GPT-6 directly with the last GPT-5 model: GPT-6 Instant starts answering questions that need web search 44% sooner on average than GPT-5.6 Instant. OpenAI notes this measures time to begin answering, not total completion time.
The original GPT-5 left ChatGPT well before GPT-6 arrived. In a January 29, 2026 post, OpenAI referred to the "previously announced retirement of GPT-5 (Instant and Thinking)" and said most usage had shifted to GPT-5.2. The same post retired GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini from ChatGPT on February 13, 2026, noting that only 0.1% of users were still choosing GPT-4o each day.
For the bigger picture on ChatGPT usage and plans, see our ChatGPT statistics and our roundup of new ChatGPT features in 2026.
Are GPT-5 Models Still Available in the API?
Yes, but only for a few more weeks. OpenAI's deprecations page says the original gpt-5, gpt-5-mini, gpt-5-nano and gpt-5-pro models shut down on December 11, 2026. OpenAI announced that deprecation on June 11, 2026.
Some GPT-5 variants are already gone. gpt-5-chat-latest and gpt-5-codex were shut down on July 23, 2026, and the GPT-5 models no longer appear on OpenAI's current models page, which now leads with GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna. The only GPT-5 family model still listed there is the specialized GPT-5.6 Cyber.
API model | Status (November 1, 2026) | Shutdown date | OpenAI's recommended replacement |
|---|---|---|---|
gpt-5 | Deprecated | December 11, 2026 | gpt-5.6-sol |
gpt-5-mini | Deprecated | December 11, 2026 | gpt-5.6-terra |
gpt-5-nano | Deprecated | December 11, 2026 | gpt-5.6-luna |
gpt-5-pro | Deprecated | December 11, 2026 | gpt-5.6-sol (pro reasoning mode) |
gpt-5.1 | Deprecated | April 1, 2027 | gpt-6-sol |
gpt-5.3-codex | Deprecated | April 1, 2027 | gpt-6-sol |
gpt-5-chat-latest, gpt-5-codex | Shut down | July 23, 2026 | gpt-5.6-sol |
Source: OpenAI API deprecations, checked November 1, 2026. OpenAI has not posted a deprecation notice for gpt-5.5.
GPT-5 Launch Statistics (August 2025)
GPT-5 launched on August 7, 2025 and began rolling out the same day to Plus, Pro, Team and Free users in ChatGPT, with Enterprise and Edu access following a week later, according to OpenAI's GPT-5 announcement. Free users who hit their usage limits were moved to GPT-5 mini.
GPT-5 reached developers on the same day. OpenAI's developer launch post listed three API sizes (gpt-5, gpt-5-mini and gpt-5-nano) and a total context length of 400,000 tokens, split into a maximum of 272,000 input tokens and 128,000 reasoning and output tokens.
GPT-5 launch metric (August 2025) | Figure | Source |
|---|---|---|
Launch date | August 7, 2025 | |
Day-one ChatGPT plans | Free, Plus, Pro, Team | |
Enterprise and Edu access | The following week | |
API context length | 400,000 tokens (272K in, 128K out) | |
Factual errors vs GPT-4o (web search on) | About 45% less likely | |
Factual errors vs o3 (thinking on) | About 80% less likely |
What Were GPT-5's Benchmark Scores?
At launch in August 2025, GPT-5 scored 94.6% on AIME 2025 math without tools, 74.9% on SWE-bench Verified, 88% on Aider Polyglot, 84.2% on MMMU and 46.2% on HealthBench Hard, according to OpenAI. These are OpenAI's own reported results.
Benchmark (August 2025) | GPT-5 score | What it measures |
|---|---|---|
AIME 2025 (no tools) | 94.6% | Competition math |
SWE-bench Verified | 74.9% (o3: 69.1%) | Fixing real GitHub issues |
Aider Polyglot | 88% | Code editing across languages |
MMMU | 84.2% | Multimodal understanding |
HealthBench Hard | 46.2% | Difficult health conversations |
GPQA Diamond (GPT-5 pro, no tools) | 88.4% | PhD-level science |
Sources: Introducing GPT-5 and Introducing GPT-5 for developers, both dated August 7, 2025. An earlier version of this page credited the 88.4% GPQA score to standard GPT-5; OpenAI reports it for GPT-5 pro with extended reasoning.
GPT-5's AIME result did not stay on top for long. By December 2025, OpenAI reported 100% on AIME 2025 for GPT-5.2 Thinking, which means the test no longer separates the strongest models, so newer launches lean on agent and computer-use benchmarks instead.
How Did GPT-5.1 Through GPT-5.6 Improve on GPT-5?
Each GPT-5.x release moved the series further toward agents and professional work rather than raw test scores. The figures below come from OpenAI's launch posts for each version.
GPT-5.2 (December 11, 2025): OpenAI reported 100.0% on AIME 2025 without tools for GPT-5.2 Thinking and GPT-5.2 Pro, up from 94.0% for GPT-5.1 Thinking (OpenAI).
GPT-5.4 (March 5, 2026): 75.0% on OSWorld-Verified computer use (GPT-5.2: 47.3%), 83.0% wins or ties on GDPval knowledge work (GPT-5.2: 70.9%) and 82.7% on BrowseComp web search (GPT-5.2: 65.8%) (OpenAI).
GPT-5.5 (April 23, 2026): pitched by OpenAI as "a new class of intelligence for real work and powering agents," with a 1 million token context window in the API. Sam Altman put API pricing at $5 per million input tokens and $30 per million output tokens (OpenAI Developer Community).
GPT-5.6 (July 9, 2026): released as a family of three models, Sol, Terra and Luna (OpenAI API changelog). Notebookcheck reported that GPT-5.6 reached ChatGPT, Codex, the API and GitHub Copilot on the same day, with Sol available on Plus and higher plans.
The Sol, Terra and Luna naming that arrived with GPT-5.6 carried straight into GPT-6, which is why the ChatGPT plan names for GPT-6 (Sol for paid, Luna for Free and Go) look familiar.
How Much Did GPT-5 Cost Compared With GPT-6?
GPT-5 launched at $1.25 per million input tokens and $10 per million output tokens, with GPT-5 mini at $0.25/$2 and GPT-5 nano at $0.05/$0.40, per OpenAI's August 2025 developer post. Those prices no longer appear on OpenAI's pricing page, which now lists the GPT-6 family.
Model | Input per 1M tokens | Output per 1M tokens | Cost for 10M input + 2M output |
|---|---|---|---|
GPT-5 (launch, Aug 2025) | $1.25 | $10 | $32.50 |
GPT-5 mini (launch) | $0.25 | $2 | $6.50 |
GPT-5 nano (launch) | $0.05 | $0.40 | $1.30 |
GPT-5.5 (Apr 2026) | $5 | $30 | $110 |
GPT-6 Astra (current) | $10 | $50 | $200 |
GPT-6.1 Sol (current) | $2 | $10 | $40 |
GPT-6.1 Terra (current) | $1 | $5 | $20 |
GPT-6 Luna (current) | $0.10 | $0.50 | $2 |
Sources: OpenAI GPT-5 developer launch, GPT-5.5 announcement thread and OpenAI API pricing (standard tier, checked November 1, 2026). Cost column is our own arithmetic at list prices, before cached-input discounts.
The worked example shows how the replacement math lands for a monthly workload of 10 million input tokens and 2 million output tokens. A team moving from GPT-5 to GPT-6.1 Terra would pay $20 instead of $32.50, about 38% less. Moving to GPT-6.1 Sol would cost $40, about 23% more than GPT-5. Moving from GPT-5 nano to GPT-6 Luna would rise from $1.30 to $2.
For current per-model rates, discounts and batch pricing, see our OpenAI API pricing guide.
Should You Still Build on GPT-5?
No new project should start on GPT-5. The original GPT-5 API models stop working on December 11, 2026, ChatGPT no longer offers them, and OpenAI's own replacement guidance points to the GPT-5.6 tiers or GPT-6.
The short version for anyone with a GPT-5 integration: the model that set these records is now a migration ticket, not a roadmap item.
A sensible order of work is to list every place your code calls gpt-5, gpt-5-mini, gpt-5-nano or gpt-5-pro, test the replacement OpenAI recommends for each, and rerun your own quality checks before December 11. Teams on gpt-5.1 or gpt-5.3-codex have until April 1, 2027.
GPT-5's statistics are still useful as history. They show where OpenAI stood in August 2025 and give a fixed baseline for judging how much later models improved, as long as each number is read with its date.
How Does the EU AI Act Apply to GPT-5?
GPT-5 launched five days after the EU AI Act's obligations for general-purpose AI models took effect on August 2, 2025, according to the European Commission. That timing put GPT-5 under the new rules from day one rather than inside the grace period for older models.
The Commission says general-purpose AI providers must publish a summary of their training content, keep technical documentation and put a copyright policy in place. Its guidance for GPAI providers adds that the Commission's enforcement powers, including fines, apply from August 2, 2026, while models placed on the market before August 2, 2025 have until August 2, 2027 to comply.
OpenAI Statistics 2026
Revenue, valuation, funding and user numbers for the company behind GPT-5 and GPT-6.
ChatGPT Statistics 2026
Usage, subscribers and plan data for the app where GPT-6 replaced GPT-5.
OpenAI API Pricing
Current per-token prices for GPT-6 Astra, Sol, Terra and Luna.
New ChatGPT Features in 2026
What changed in ChatGPT this year, including the move to GPT-6.
Frequently Asked Questions
Is GPT-5 still the latest OpenAI model?
No. GPT-6 replaced the GPT-5.x models in ChatGPT in October 2026, with GPT-6 Sol rolling out to paid plans from October 7 and GPT-6 Luna to Free and Go users from October 8. The last GPT-5 release was the GPT-5.6 family in July 2026.
When was GPT-5 released?
OpenAI released GPT-5 on August 7, 2025. It rolled out that day to Free, Plus, Pro and Team users in ChatGPT and to developers in the API, with Enterprise and Edu access the following week.
What were GPT-5's benchmark scores?
At launch in August 2025, OpenAI reported 94.6% on AIME 2025 without tools, 74.9% on SWE-bench Verified, 88% on Aider Polyglot, 84.2% on MMMU and 46.2% on HealthBench Hard. GPT-5 pro scored 88.4% on GPQA Diamond without tools.
Can you still use GPT-5 in the OpenAI API?
Yes, until December 11, 2026. OpenAI deprecated gpt-5, gpt-5-mini, gpt-5-nano and gpt-5-pro on June 11, 2026 and recommends the GPT-5.6 Sol, Terra and Luna models as replacements; gpt-5.1 runs until April 1, 2027.
How much did GPT-5 cost in the API?
GPT-5 launched at $1.25 per million input tokens and $10 per million output tokens, with GPT-5 mini at $0.25/$2 and GPT-5 nano at $0.05/$0.40. For comparison, OpenAI's current GPT-6.1 Sol lists at $2/$10.
The Bottom Line
GPT-5 was OpenAI's flagship from August 2025 until the GPT-5.x updates and then GPT-6 took over. Its launch numbers (94.6% on AIME 2025, 74.9% on SWE-bench Verified, $1.25/$10 API pricing) are now historical markers rather than current specs.
For anyone using ChatGPT today, the model is GPT-6 Sol or GPT-6 Luna depending on plan. For developers, the deadline that matters is December 11, 2026, when the original GPT-5 API models shut down.
Written by the AI Business Weekly Editorial Team. Reviewed, edited and published by Sameer Khan. This article was AI-assisted, then reviewed by Sameer Khan before publishing.
