GPT-5.1

GPT-5.1 is a family of five large language models from OpenAI's GPT-5 series, released in two waves in November 2025: the first three models (Instant, Thinking, Codex-Mini) on November 12, and two more (Codex-Max and Pro) one week later on November 19. The update stays within the GPT-5 generation but delivers meaningful improvements in intelligence, communication style, and personalisation.

Family variants

GPT-5.1 Instant — the most-used model, warmer and more conversational by default, better at following instructions. For the first time in an 'Instant' variant, adaptive reasoning appears: the model itself decides when to 'think' before responding to harder questions. In the API available as gpt-5.1-chat-latest.

GPT-5.1 Thinking — the advanced reasoning model, now more efficient and easier to understand in everyday use. Dynamically adapts thinking time to the query: about twice as fast on the easiest tasks and twice as slow on the hardest ones compared to GPT-5 Thinking. In the API as gpt-5.1.

GPT-5.1 Auto — a router that directs a query to the most appropriate variant without the user having to pick a model.

GPT-5.1 Pro — the top tier for professional and complex tasks, replacing GPT-5 Pro from November 19, 2025.

GPT-5.1-Codex-Mini — a cost-effective variant for coding and agentic tasks. Released November 12.

GPT-5.1-Codex-Max — a next-generation agentic coding model. It has internally documented completion of tasks lasting up to 24 hours. Released November 19.

Key capabilities

• Adaptive reasoning — dynamic inference budget: faster for easy questions, more cycles for hard ones.

• Long context of 400,000 input tokens and up to 128,000 output tokens.

• New agentic tools: apply_patch (safe file modifications) and shell (executing system commands) — key for the Codex variants.

• Extended prompt caching — lowers the cost of applications operating on long contexts.

• 8 personality options: Default, Friendly, Efficient, Professional, Candid, Quirky, Nerdy, Cynical. Plus experimental fine-grained tuning (concise, warm, scannable, emoji frequency) and the ability for ChatGPT to proactively offer to update preferences during a conversation.

• Significantly better instruction following than GPT-5 — e.g. reliably sticking to constraints like 'reply in six words'.

Availability and pricing

Available to all ChatGPT users, rollout from paid (Pro, Plus, Go, Business) to free/logged-out. Enterprise and Edu have a 7-day early access toggle. GPT-5 (Instant and Thinking) remains in the 'legacy models dropdown' for 3 months for comparison. API pricing: USD 1.25 per 1M input tokens, USD 10 per 1M output tokens.

Reception and positioning

Reviewers describe GPT-5.1 as an 'incremental but meaningful improvement' (TechRadar). Tom's Guide praises mathematical reasoning and coding while noting that Gemini 3 performed better in a direct qualitative comparison. Inside OpenAI's lineup, GPT-5.1 is a standard mid-cycle update to the GPT-5 generation, preparing the ground for GPT-5.2 and further iterations within the family.