A general-purpose assistant for chat, research, coding, files, images, voice and end-to-end knowledge work.
ChatGPT is the product. GPT-6 Sol powers paid plans and GPT-6 Luna powers Free/Go tiers. GPT-6 Astra has a separate Aivorix model dossier. A model listing does not establish access on every ChatGPT plan.
ChatGPT began as a free research preview built on GPT-3.5 and has grown into a multi-tool workspace rather than a text-only chatbot.
The GPT-6 family separates flagship professional quality (Sol) and fast low-cost conversational work (Luna). Intelligent UI allows responses to dynamically render interactive visual elements directly in chat.
Product features and underlying models are documented separately. Roadmap items include only provider-announced work; rumors are excluded.
Selected milestones that materially changed the product.
OpenAI transitioned ChatGPT tiers to GPT-6 Sol and Luna, adding native interactive UI rendering and Dots persistent agents.
OpenAI released the conversational GPT-3.5 experience to collect feedback.
ChatGPT added stronger reasoning, then moved toward a natively multimodal, faster experience across text, vision and voice.
A unified system combined fast responses with deeper reasoning and improved instruction following, coding and factual reliability.
The family added durable capability tiers, configurable effort, stronger agent work and improved price-performance.
The comparison uses provider-published capabilities and named evaluations, not a generic marketing score.
| Dimension | Previous | Current | Measured difference |
|---|---|---|---|
| Professional agent tasks | 46.9% on Agents' Last Exam | 52.7% | +5.8 percentage points (about 12.4% relative) |
| Research workflow capability | GPT-5.5 baseline | GPT-5.6 Sol | +16.2 points on OpenAI's internal aggregate RSI evaluation |
| Tool orchestration | Tool calls handled mainly through the host loop | Can write small in-memory programs and coordinate parallel agents | Fewer round trips for multi-step workflows |
| Product structure | Single generation with model variants | Sol / Terra / Luna capability tiers plus effort controls | More explicit quality, speed and cost choice |
Capability changes by task. The fairest published comparison is benchmark-by-benchmark; adoption or token volume is not treated as intelligence.
52.7% versus 46.9% for GPT-5.5 on long-running professional workflows.
Check evidenceSol used less than half the time and output tokens of Claude Fable 5 on OpenAI's cited Coding Agent Index comparison.
Check evidenceSol Max came within one point of Fable 5 on the AA Intelligence Index while finishing in 61% less time.
Check evidenceWhat happens behind the button, where it helps and what to verify.
How it works: Plans a research path, searches multiple sources, reads them and produces a cited synthesis instead of relying only on model memory.
Best use: Market scans, literature reviews and evidence-heavy questions
Watch: Always open the cited sources; a citation can still be incomplete or misinterpreted.
How it works: Parses documents or tables and can use a code-execution environment to calculate, transform and visualize data.
Best use: PDF review, spreadsheets, charts and structured extraction
Watch: Check formulas, units and sampled rows before using results operationally.
How it works: Higher effort allocates more reasoning and, at the top end, can coordinate multiple workstreams; lower effort prioritizes speed.
Best use: Matching response cost and latency to task difficulty
How it works: Accepts speech and images as context and returns conversational or visual assistance through supported clients.
Best use: Hands-free help, visual inspection and accessibility
Watch: Do not treat visual or voice output as a substitute for expert inspection in high-stakes cases.
Status reflects the provider's latest public wording. “No public date” means there is no verified launch schedule.
Multi-agent support launched in beta for supported workflows; availability and behavior may change during rollout.
Official announcementOpenAI is separately tuning GPT-5.6 for everyday ChatGPT conversations while broadening Luna access for Free users.
Official announcementOpenAI has not published a release date or guaranteed feature list for a successor to GPT-5.6.
Provider announcements, help documentation and system information.
OpenAI · Nov 30, 2022
OpenAI · Jul 9, 2026
OpenAI · Aug 6, 2026
Last editorial verification: 2026-11-02. Prices, access limits, preview status and model routing can change after this date.