Introducing GPT‑6.2 Astra and Sol
2026-11-18 · OpenAI
Introducing GPT‑6.2 Astra and Sol
The announcement introduces GPT‑6.2 Astra and Sol, describing them as delivering better professional outputs, coding, computer use, and efficiency at lower cost across ChatGPT Work, Codex, and the API.
What the announcement contains
The published text is brief: a title followed by a single descriptive sentence. It names two models — Astra and Sol — under a shared "GPT‑6.2" version label, and frames the release around four areas of improvement plus a cost argument. No further paragraphs, data, or explanation accompany the statement.
The four areas of improvement
- Professional outputs. The models are said to produce better results for professional work.
- Coding. Programming-related capability is described as improved.
- Computer use. Computer use is listed among the areas that got better.
- Efficiency at lower cost. The sentence pairs improved efficiency with lower cost, presenting the update as both a capability change and an economic one.
Where the improvements apply
The announcement states that the gains apply "across" three surfaces:
- ChatGPT Work — the work-oriented ChatGPT offering named in the text.
- Codex — the Codex product named in the text.
- API — developer access.
Because the sentence uses "across," the claim covers all three surfaces rather than a single product. The release is therefore presented as a coordinated update spanning consumer-facing, coding, and developer channels.
Two names, one version
Astra and Sol appear together under GPT‑6.2. The announcement does not explain whether they differ in size, speed, modality support, or intended workload, nor whether one supersedes the other or both are offered in parallel.
What is not in the announcement
- Benchmark results or comparisons with earlier models.
- Pricing, rate limits, or the magnitude of the cost reduction.
- Technical specifications such as context length or supported modalities.
- Release dates, staged rollout plans, or regional availability.
- Safety evaluation results or system-card details.
- Guidance on choosing between Astra and Sol.
Reading the release
The verifiable content is limited to direction and scope: which capabilities are claimed to improve, and which products those claims cover. Everything else — figures, timelines, specifications — is absent from the source text.
The four capability claims should therefore be read as the publisher's characterization of its own models rather than as measured outcomes. Anyone assessing the release will need the accompanying technical documentation, benchmark data, and pricing information before drawing conclusions about real-world performance or cost impact.