Skip to content
TechPP

Festive sale: the best deals in one sheet. See the deals

HomeUpdate TrackerOpenAI GPT Updates

OpenAI GPT Model Update Tracker

OpenAI first announced ChatGPT in 2023, and it created a storm around the world. Since then, AI models for regular users have been made available, and the latest models have been phenomenal at what they do.

Although models in the beginning focused on multiple capabilities, including image and video generation, the transition slowly happened to coding. Now every model is competing for the top coding benchmarks, including OpenAI.

OpenAI is going through a big shift, focusing on coding and its Codex model, and bringing a lot of changes recently. They announced a super app and rolled out a couple of updates to the Codex app, which is also available for Mac and Windows.

OpenAI's current GPT-6 family includes Astra, Sol, and Luna. GPT-6 Astra arrived on September 3, 2026, followed by GPT-6 Sol and Luna on September 22 and GPT-6.1 Sol on September 29.

In the update trackers, we track all the latest updates to the GPT models. As always, you can bookmark this post for the latest updates.

OpenAI Logo

GPT Model Releases

GPT-6.1 Sol

September 29, 2026
  • NEW
    Release:OpenAI released GPT-6.1 Sol, a point update to GPT-6 Sol that brings it close to GPT-6 Astra on coding, computer use and professional work
  • PRICE
    Cost:Keeps GPT-6 Sol's $2 per million input tokens and $10 per million output tokens, while cached input drops by half to $0.10 per million
  • IMPROVED
    Cost:Costs one-fifth of GPT-6 Astra's per-token price
  • CAPABILITY
    Context:Keeps the 1.05M-token context window and 128K max output
  • NEW
    Availability:Available in the OpenAI API as gpt-6.1-sol and in Codex

GPT-6 Sol and Luna

September 22, 2026
  • NEW
    Release:OpenAI released GPT-6 Sol and GPT-6 Luna, faster and cheaper models built on the advances behind GPT-6 Astra
  • PRICE
    Cost:Sol costs $2 per million input tokens and $10 per million output tokens; Luna costs $0.10 and $0.50
  • CAPABILITY
    Context:Both offer a 1.05M-token context window and 128K max output
  • NEW
    Availability:Available in ChatGPT Work, Codex, the OpenAI API and the major clouds, but not yet in regular ChatGPT chat

GPT-6 Astra

September 3, 2026
  • NEW
    Release:OpenAI released GPT-6 Astra, its new flagship model
  • FEATURE
    Safety:The first OpenAI model rated Critical for cybersecurity, so access started with companies in the Daybreak programme before ChatGPT Plus, Pro, Business and Enterprise, the API and AWS followed
  • CAPABILITY
    Agents:Operates software through the screen, with computer use, asynchronous tool calls, multi-agent workflows and mid-task steering
  • FEATURE
    Reasoning:Supports five reasoning levels, from low to max
  • BENCHMARK
    Science:Scores 96.0% on GPQA Diamond and 97.6% on FrontierMath Tier 4 v2
  • BENCHMARK
    Coding:Scores 74.1% on DeepSWE v1.1 and 57.9% on Terminal-Bench 4.0
  • PRICE
    Cost:Priced at $10 per million input tokens and $50 per million output tokens
  • CAPABILITY
    Context:Offers a 1.05M-token context window, 128K max output and an April 30, 2026 knowledge cutoff

GPT-5.6 Sol API price reduction

August 21, 2026
  • PRICE
    API:OpenAI reduced GPT-5.6 Sol API and credit pricing by more than 20% for a three-month period

GPT-5.6 ChatGPT updates

August 6, 2026
  • UPDATE
    ChatGPT:OpenAI updated GPT-5.6 Sol for chats
  • UPDATE
    Free:GPT-5.6 Luna became the unlimited default model for Free users

GPT-5.6 Sol, Terra and Luna general availability

July 9, 2026
  • NEW
    Release:GPT-5.6 Sol, Terra, and Luna became generally available across ChatGPT, Codex, and the OpenAI API
  • NEW
    Models:Sol is the frontier model, Terra is the balanced everyday model, and Luna is the fast, lower-cost model

GPT-5.6 Sol, Terra & Luna (limited preview)

June 26, 2026
  • NEW
    Models:OpenAI previewed the GPT-5.6 family: Sol (flagship), Terra (balanced everyday model), and Luna (fast, affordable), roughly two months after GPT-5.5
  • IMPROVED
    Agents:GPT-5.6 shows improved agentic capabilities in coding, biology, and cybersecurity
  • UPDATE
    Availability:Rollout is limited to a small group of trusted partners at the request of the U.S. government; broad general availability is planned in the coming weeks
  • PRICE
    Cost:Per million tokens: Sol $5 input / $30 output, Terra $2.50 / $15, Luna $1 / $6

GPT-5.2 retired, GPT-5.6 incoming

June 16, 2026
  • LIMITATION
    Retirements:GPT-5.2 is no longer available in ChatGPT as of June 12; existing GPT-5.2 conversations automatically continue on the matching GPT-5.5 model
  • NEW
    Upcoming:GPT-5.6 is expected by late June 2026; OpenAI's chief scientist describes it as a meaningful improvement over GPT-5.5

ChatGPT & model updates (June 2026)

June 9, 2026
  • FEATURE
    Personalization:Personalization is rolling out to ChatGPT Go and Free; Free-tier responses now draw from a reduced set of your past chats
  • NEW
    Codex:New GPT-5-Codex-Mini model added to the Codex CLI and IDE extension as a smaller, more cost-effective version of GPT-5-Codex
  • LIMITATION
    Retirements:GPT-4.5 retires from ChatGPT on June 27, 2026 (30-day sunset); OpenAI o3 retires on August 26, 2026 (90-day sunset)

GPT-5.5 Instant

May 6, 2026
  • FEATURE
    Release:GPT-5.5 Instant is rolling out as the new default ChatGPT model, replacing GPT-5.3 Instant for all ChatGPT users.
  • FEATURE
    ChatGPT:Paid users can continue using GPT-5.3 Instant for three months through model configuration settings before it is retired.
  • IMPROVED
    Core Quality:Better everyday answers across accuracy, clarity, and conciseness.
  • IMPROVED
    Image Understanding:Better at analyzing and understanding images.
  • IMPROVED
    STEM:Better performance on science, technology, engineering, and math questions.
  • IMPROVED
    Web Search:Better at deciding when ChatGPT should search the web.
  • IMPROVED
    Context:For Plus and Pro users on web, GPT-5.5 Instant can better use context from past chats, files, and connected Gmail to make responses more relevant.
  • FEATURE
    Tools:Supports all current ChatGPT tools, including web search, data analysis, image analysis, file analysis, canvas, image generation, memory, and custom instructions.
  • SAFETY
    Safety:First Instant model OpenAI is treating as High capability in Cybersecurity and Biological & Chemical Preparedness categories, with added safeguards.
  • FEATURE
    Enterprise:Disabled by default for ChatGPT Enterprise workspaces. Admins and owners can enable it in workspace settings.
  • LIMITATION
    Healthcare:Not available for ChatGPT for Healthcare workspaces.

GPT-5.5

Apr 23, 2026
  • NEW
    Release:New top GPT model for complex real-world work, positioned as more practical for hard tasks due to lower latency
  • IMPROVED
    Capabilities:Better at coding, research, data analysis, document-heavy work, and agent-style tasks across tools
  • IMPROVED
    Agentic:Better at understanding complex goals early, asking for less guidance, checking its work, and completing tasks
  • FEATURE
    ChatGPT:GPT-5.5 Thinking is tuned for faster help on harder problems with smarter and more concise answers
  • FEATURE
    ChatGPT:GPT-5.5 Pro is positioned as a stronger version for more demanding work with better structure and accuracy
  • BENCHMARK
    Knowledge Work:Reached 84.9% on GDPval for well-specified knowledge work across 44 occupations
  • BENCHMARK
    Agents:Reached 78.7% on OSWorld-Verified for operating real computer environments
  • BENCHMARK
    Performance:98.0% on Tau2-bench Telecom, 60.0% on FinanceAgent, 88.5% on internal investment-banking tasks, and 54.1% on OfficeQA Pro
  • IMPROVED
    Research:Better at multi-stage scientific and technical research workflows, including testing assumptions and interpreting results
  • BENCHMARK
    Science:Improved over GPT-5.4 on GeneBench and posted leading performance on BixBench
  • CAPABILITY
    Discovery:Internal setup helped discover a new proof related to Ramsey numbers, later verified in Lean
  • NEW
    Safety:Launched alongside a dedicated System Card and a Bio Bug Bounty with up to $25,000 in rewards
  • NEW
    Availability:Available in ChatGPT and Codex starting April 23, 2026

GPT-5.4 nano

Mar 17, 2026
  • NEW
    Positioning:Cheapest GPT-5.4-class model
  • CAPABILITY
    Focus:Built for classification, ranking, extraction, and simple high-volume tasks
  • IMPROVED
    Performance:Faster small-model option for API-heavy workflows
  • FEATURE
    Context:Comes with a 400K context window
  • NEW
    Availability:API-only at launch

GPT-5.4 mini

Mar 17, 2026
  • NEW
    Release:Faster and cheaper GPT-5.4-class model
  • IMPROVED
    Focus:Optimized for coding, tool use, multimodal reasoning, and high-volume workloads
  • BENCHMARK
    Speed:More than 2x faster than GPT-5 mini
  • BENCHMARK
    Performance:Gets close to full GPT-5.4 on several evals including SWE-Bench Pro and OSWorld-Verified
  • CAPABILITY
    Use Cases:Built for subagents, fast iteration, and cheaper production usage
  • FEATURE
    Context:Comes with a 400K context window
  • NEW
    Availability:Available in API, Codex, and ChatGPT

GPT-5.4

Mar 5, 2026
  • NEW
    Release:Current top GPT model in OpenAI’s public docs
  • IMPROVED
    Intelligence:Best intelligence at scale for agentic, coding, and professional workflows
  • IMPROVED
    Performance:Better than GPT-5.2 on knowledge work, spreadsheets, documents, and presentations
  • BENCHMARK
    Performance:Reached 83.0% on GDPval versus 70.9% for GPT-5.2
  • BENCHMARK
    Agents:Reached 75.0% on OSWorld-Verified, above GPT-5.2 at 47.3% and above the human result listed there at 72.4%
  • FEATURE
    Control:Supports reasoning effort up to xhigh
  • FEATURE
    Context:Comes with a 1M context window

GPT-5.1

Nov 12, 2025
  • IMPROVED
    General:Smarter and more conversational GPT-5 update
  • IMPROVED
    Balance:Better balance of intelligence and speed for coding and agentic work
  • FEATURE
    Reasoning:Added a new 'none' reasoning mode for lower-latency responses
  • IMPROVED
    Calibration:Better calibrated to prompt difficulty
  • IMPROVED
    Control:More steerable in personality, tone, and formatting
  • FEATURE
    Context:Uses a 400K context window

GPT-5

Aug 7, 2025
  • NEW
    Release:New default model in ChatGPT at launch
  • NEW
    Availability:Replaced GPT-4o, GPT-4.1, GPT-4.5, o3, and o4-mini for signed-in ChatGPT users
  • CAPABILITY
    Reasoning:Handles reasoning automatically when needed
  • NEW
    Positioning:Positioned as a more unified model instead of making users choose between many separate workflows

GPT-4.1 mini

Apr 14, 2025
  • NEW
    Release:Smaller and faster version of GPT-4.1
  • IMPROVED
    Performance:Better than GPT-4o mini in instruction following, coding, and general intelligence
  • BENCHMARK
    Performance:Matches or beats GPT-4o on some benchmarks
  • IMPROVED
    Speed:Cuts latency by nearly half
  • PRICING
    Cost:Cuts cost by 83% versus GPT-4o
  • FEATURE
    Context:Comes with a 1M token context window
  • NEW
    Availability:Replaced GPT-4o mini in ChatGPT on May 14, 2025

GPT-4.1

Apr 14, 2025
  • NEW
    Positioning:Strongest non-reasoning GPT model at launch
  • IMPROVED
    Performance:Better at instruction following and tool calling than GPT-4o
  • FEATURE
    Context:Comes with a 1M token context window
  • IMPROVED
    Speed:Lower latency than reasoning-heavy models
  • NEW
    Positioning:Better choice than GPT-4.5 for many practical API use cases

GPT-4.5

Feb 27, 2025
  • NEW
    Release:Research preview of OpenAI’s strongest GPT chat model at launch
  • IMPROVED
    Training:Scaled up pre-training and post-training
  • IMPROVED
    Intelligence:Better at recognizing patterns, drawing connections, and generating creative insights
  • IMPROVED
    Conversation:Stronger for chat quality and natural conversation
  • NEW
    Positioning:Later positioned for deprecation in favor of GPT-4.1 for many API uses

GPT-4o

May 13, 2024
  • NEW
    Release:New flagship model with GPT-4-level intelligence but much faster speed
  • IMPROVED
    Multimodal:Stronger across text, vision, and audio
  • NEW
    Architecture:Built as a true omni model instead of stitching separate models together
  • NEW
    Availability:Pushed advanced capability to free ChatGPT users too

GPT-4 Turbo

Nov 6, 2023
  • IMPROVED
    Release:Faster and cheaper GPT-4-class model
  • FEATURE
    Context:Added 128K context window
  • IMPROVED
    Tools:Improved tool use and developer features
  • NEW
    Capabilities:Made GPT-4-level capability easier to use in production

GPT-4

Mar 14, 2023
  • NEW
    Release:First GPT model described as multimodal, with image and text input support
  • IMPROVED
    Intelligence:Large jump in reasoning, exams, and professional benchmark performance
  • BENCHMARK
    Exams:Scored around the top 10% on a simulated bar exam, while GPT-3.5 was near the bottom 10%
  • NEW
    Positioning:Became the main high-end GPT model for complex tasks

GPT-3.5 Turbo

Mar 1, 2023
  • IMPROVED
    Release:Optimized GPT-3.5 model for developers
  • IMPROVED
    Performance:Lower cost and faster than earlier GPT-3.5 variants
  • NEW
    Usage:Became the practical default for many production chat apps
  • NEW
    Migration:Helped move ChatGPT-style workflows into the API

GPT-3.5

Nov 30, 2022
  • IMPROVED
    Chat:Much better for chat than GPT-3
  • NEW
    Training:Fine-tuned for dialogue
  • IMPROVED
    Performance:Better at following instructions and staying conversational
  • NEW
    Release:Became the model family behind the first public ChatGPT launch

GPT-3

May 28, 2020
  • IMPROVED
    Scale:Big leap in scale and few-shot learning
  • IMPROVED
    Prompts:Made prompt-based usage much more practical
  • CAPABILITY
    Generation:Can solve many tasks from examples in the prompt instead of fine-tuning
  • NEW
    Use Cases:Turned GPT into a general-purpose text engine for apps and APIs

GPT-2

Feb 14, 2019
  • IMPROVED
    Generation:Much larger jump in text generation quality
  • CAPABILITY
    Generation:Can generate long, coherent paragraphs without task-specific training
  • IMPROVED
    Zero-shot:Stronger zero-shot performance across reading comprehension, translation, summarization, and question answering
  • NEW
    Release:Released in stages due to safety concerns. Final 1.5B parameter version was fully released on Nov 5, 2019

GPT-1

Jun 11, 2018
  • NEW
    Release:First GPT model release
  • NEW
    Training:Introduced generative pre-training plus task-specific fine-tuning
  • CAPABILITY
    Transfer:Showed that one large language model could transfer across multiple NLP tasks
  • NEW
    Foundation:Set the base for the full GPT line

Was this article helpful?

Comments

0

Share your thoughts, questions, or feedback about this guide.

Leave a Comment

0/2000