OpenAI's GPT-5.x line runs from GPT-5.1 to GPT-5.6 Sol
Between GPT-5 and GPT-6 OpenAI shipped reasoning-capable models from GPT-5.1 to GPT-5.6 Sol.
- Date
- 12 November 2025
- Who
- OpenAI
- Confidence
- Medium (dates from OpenAI's API changelog and Wikipedia pointers; OpenAI's launc
- Deep dive
- Reasoning II, from o1 and o3 to DeepSeek-R1 and the labs that replicated them
Tier: Supporting · Significance: 3/5 · Org(s): OpenAI · Confidence: Medium (dates from OpenAI's API changelog and Wikipedia pointers; OpenAI's launch posts returned HTTP 403) Between GPT-5 and GPT-6 OpenAI shipped reasoning-capable models from GPT-5.1 to GPT-5.6 Sol. GPT-5.1 (ChatGPT 2025-11-12 per Wikipedia; API changelog 2025-11-13) came as Instant and Thinking variants tuned for steerability and for faster answers when less thinking is needed. GPT-5.2 (2025-12-11) added Instant, Thinking and Pro variants, an xhigh reasoning effort, concise reasoning summaries and context compaction in the API; it arrived about three weeks after Google's Gemini 3 Pro amid press reports of an internal "code red," which OpenAI's Fidji Simo disputed by saying the model had been in the works for months. GPT-5.3-Codex followed on 2026-02-24, GPT-5.4 on 2026-03-05 (tool search, built-in computer use, a 1M-token context), GPT-5.5 on 2026-04-23 (API 04-24; company-reported FrontierMath Tier 1-3 51.7% and Tier 4 35.4%, per Wikipedia citing OpenAI) and GPT-5.6 Sol on 2026-07-09 (OpenAI API changelog, Wikipedia on GPT-5.1, 5.2 and 5.5, as pointers). I infer from a named model roughly every quarter, with an effort ladder in place of separate reasoning models, that reasoning had become a default dial, the 2026 pattern in B08-51. I did not find a source for the claim, seen in secondary coverage, that GPT-5.5 was OpenAI's first fully retrained base model since GPT-4.5, and Wikipedia's GPT-5.5 article does not make it. Sources: API changelog · Wikipedia pointers above