GPT-5.6 Sol — Specs, Pricing & Limits

GPT-5.6 Sol is the model the Codex CLI actually defaults to — not the Codex-branded model, which is a separate entry in this reference and the subject of this site's most-read comparison. The unsuffixed "GPT-5.6" name is documented as an alias that routes here, which is worth knowing since it means the alias and the explicit id are, for pricing and behaviour purposes, the same model.

Modality-wise, Sol handles text in and out and accepts image input, but does not support audio or video — worth checking before assuming a multimodal workflow is fully covered.

Its published long-context repricing rule is the one that governs the site's own cross-provider and token-cost tools by default, since Sol is the model a fresh Codex CLI session actually runs unless you switch it. If your workload sends a large accumulated context — a real risk in a long agentic session — Sol's context window gives real headroom before the repricing threshold, but the threshold itself is measured on input tokens specifically and applies to the whole request once crossed, not a marginal slice of it.

The table below is computed directly from this site's facts module: pricing across every published service tier, the long-context repricing rule, and the published rate-limit tiers.