OpenAI Mini and Nano Models Compared: Specs, Prices and Shutdown Dates Side by Side

OpenAI's mini and nano models are its smaller, lower-cost GPT models for high-volume work, and in November 2026 most of them are deprecated with fixed shutdown dates. This page compares every mini and nano text model on OpenAI's API documentation as of 15 November 2026, together with GPT-6 Luna and GPT-6.1 Terra, the models OpenAI names as their replacements. It covers API IDs, release dates, context windows, knowledge cutoffs, prices, tools and shutdown dates, using identical fields for every model. It is written for developers, architects and buyers who run classification, extraction, routing or sub-agent workloads on small models and need to plan a migration. Models are listed by release date, newest first, an objective order that implies no ranking.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It uses OpenAI's public documentation as of November 2026. Model details change, so confirm against OpenAI's current documentation.

Key takeaways

• Two replacement models are current. OpenAI's deprecations page names gpt-6-luna as the replacement for gpt-5-nano, gpt-5.4-nano, GPT-4o mini and GPT-4.1 mini, and gpt-6.1-terra as the replacement for gpt-5-mini.
• GPT-5.4 mini is the only mini model not on a shutdown list. On the pages checked, gpt-5.4-mini has no deprecation entry, while gpt-5.4-nano shuts down on 1 April 2027.
• GPT-5 mini and nano shut down on 11 December 2026. The notice date was 11 June 2026, and GPT-4o mini and GPT-4.1 mini follow on 14 January 2027.
• Context runs from 128,000 to 1,050,000 tokens. GPT-4o mini lists 128,000 tokens, the GPT-5 and GPT-5.4 small models list 400,000, and GPT-6 Luna and GPT-6.1 Terra list 1,050,000.
• Input prices run from $0.05 to $1.00 per million tokens. GPT-5 nano lists $0.05 input and $0.40 output; GPT-6 Luna lists $0.10 and $0.50; GPT-6.1 Terra lists $1.00 and $5.00.

What are OpenAI mini and nano models?

OpenAI has used the suffixes "mini" and "nano" for smaller versions of a model generation: GPT-4o mini, GPT-4.1 mini and nano, GPT-5 mini and nano, GPT-5.4 mini and nano, plus the reasoning models o1-mini, o3-mini and o4-mini. OpenAI describes GPT-5.4 mini as "our strongest mini model yet for coding, computer use, and subagents" and GPT-5.4 nano as "our cheapest GPT-5.4-class model for simple high-volume tasks", naming classification, data extraction, ranking and sub-agents as nano use cases.

From GPT-5.6 onwards OpenAI names its tiers Sol, Terra and Luna instead. The GPT-6 Luna page describes it as "our most efficient model for focused, high-volume tasks", and the GPT-6.1 Terra page calls it a fast, efficient model for visual reasoning, precise computer use and clear writing. OpenAI does not label Luna or Terra as mini or nano; they appear here because the deprecations page names them as the replacements for the mini and nano models. See the LLM Naming and Versioning Explained for how suffixes, aliases and snapshots work in general, and the OpenAI Models hub for the full line-up.

Every model at a glance

As of 2026-11-15. Status is the label on OpenAI's model page or deprecations page.

Already shut down on 23 October 2026: gpt-4.1-nano, o3-mini and o4-mini. o1-mini shut down on 27 October 2025. These are listed in the deprecations section below and are not callable.

When was each model released?

As of 2026-11-15. Dates from snapshot identifiers and the OpenAI API changelog.

Dates for the 2024 and 2025 models are the dates in their snapshot identifiers. For every OpenAI launch across 2026, see the LLM Release Tracker.

How much does each model cost?

Standard prices in USD per million tokens, as of 2026-11-15. Sources: each model page and the OpenAI pricing page.

On the Sol tier, which is outside the scope of this page: OpenAI's pricing page states "GPT-5.6 Sol's promotional pricing is available until November 30, 2026", while its price tables list the model ID gpt-6.1-sol and contain no gpt-5.6-sol row (checked 2026-11-15). The page does not say which listed ID the note refers to. The deprecations page names gpt-5.6-sol as the replacement for o3-mini.

The pricing page's Flex/Batch table lists GPT-6 Luna and GPT-6.1 Terra at half their Standard rates; the older mini and nano models do not appear in the main pricing tables, so their batch rates are not shown. The pricing page states that regional processing (data residency) endpoints carry a 10% uplift for models released on or after 5 March 2026, and FedRAMP endpoints a 10% uplift. Moving from GPT-5 nano to GPT-6 Luna doubles the input price ($0.05 to $0.10) and raises output from $0.40 to $0.50; moving from GPT-5 mini to GPT-6.1 Terra raises input from $0.25 to $1.00. For how cached input, cache writes and long-prompt multipliers change a bill, see Model Pricing Explained and the LLM Pricing Tracker.

What changed between models?

Only changes OpenAI documents on its model pages, deprecations page and API changelog are listed.

• GPT-4o mini to GPT-4.1 mini: the model pages list the context window rising from 128,000 to 1,047,576 tokens, maximum output from 16,384 to 32,768 tokens, and knowledge cutoff moving from 2023-10-01 to 2024-06-01. Both list fine-tuning as supported.
• GPT-4.1 mini to GPT-5 mini and nano: the GPT-5 small models list reasoning token support, a 400,000-token context window and 128,000 maximum output, and fine-tuning is listed as not supported.
• GPT-5 to GPT-5.4 mini and nano: the changelog entry for 2026-03-17 states that GPT-5.4 mini "brings GPT-5.4-class capabilities to a faster, more efficient model for high-volume workloads" and supports tool search, built-in computer use and compaction, while GPT-5.4 nano supports compaction but not tool search or computer use. The model pages list reasoning effort from none (default) to xhigh and a knowledge cutoff of 2025-08-31.
• GPT-5.4 to the GPT-5.6 tiers: the changelog entry for 2026-07-09 introduces Sol, Terra and Luna and adds Programmatic Tool Calling, explicit prompt caching controls, persisted reasoning, max reasoning effort and a multi-agent beta. The entry for 2026-07-30 states "GPT-5.6 Luna costs 80% less, while GPT-5.6 Terra costs 20% less".
• GPT-6 Luna: the changelog entry for 2026-09-22 describes GPT-6 Sol and Luna as reasoning models that "accept text and image inputs and generate text through the Responses and Chat Completions APIs". The model page lists a 1,050,000-token context window and a default reasoning effort of medium. The changelog records the Decisions API as generally available with gpt-6-luna on 2026-10-22, with up to 400 questions per request.
• GPT-6.1 Terra: the changelog entry for 2026-10-26 describes it as "a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing". The model page lists a knowledge cutoff of 2026-06-01, the latest of the models on this page.

For background on reasoning effort settings, see Reasoning Models Explained.

What workloads does OpenAI position each model for?

As of 2026-11-15. This section records the provider's own positioning wording from each model page; it is not an assessment.

What is the context window of each model?

As of 2026-11-15, from each model's page on developers.openai.com.

None of these models accepts audio or video input on the API model pages; OpenAI lists audio and video as not supported for each. The default reasoning effort differs: GPT-5.4 mini and nano default to none, while GPT-6 Luna and GPT-6.1 Terra default to medium, so a like-for-like migration should set the effort explicitly rather than rely on the default. The GPT-6 Luna page also notes that Chat Completions supports function calling only when reasoning_effort is none, and points to the Responses API for built-in tools. See Context Windows Explained and Knowledge Cutoff Explained.

Capabilities and tools compared

As of 2026-11-15, from the Features and Tools sections of each model page.

On the Responses API, GPT-6 Luna, GPT-6.1 Terra and GPT-5.4 mini list web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP and tool search. GPT-5.4 nano lists the same set without computer use and tool search. The rate-limit tables on the GPT-6 Luna, GPT-5.4 mini and GPT-5.4 nano pages are identical: not supported on the Free tier, then 5,000 requests and 2,000,000 tokens per minute on Build, rising to 30,000 requests and 180,000,000 tokens per minute on Grow. Teams that fine-tuned GPT-4o mini or GPT-4.1 mini should note that fine-tuning is listed as not supported on GPT-6 Luna, the stated replacement for both.

Where is each model available?

As of 2026-11-15, from each model page.

• OpenAI API: every model on this page lists the Chat Completions and Responses endpoints, and the model pages also list the Batch endpoint.
• Amazon Bedrock: the GPT-6 Luna and GPT-6.1 Terra pages list Bedrock IDs openai.gpt-6-luna and openai.gpt-6.1-terra.
• Google Cloud: the same two pages list gpt-6-luna and gpt-6.1-terra on Google Cloud as private preview.
• Data residency: the GPT-6 Luna page states that EU data residency is available with Standard, Fast, Flex and Batch processing.

Model cards and system cards

As of 2026-11-15. None of the eight OpenAI model pages checked links to a system card, model card or technical report.

Benchmarks

The OpenAI model pages checked for these eight models give qualitative ratings (for example "Reasoning: High" and "Speed: Fast" for GPT-6 Luna) rather than benchmark scores. Provider-reported benchmark figures for these models are not stated by OpenAI on the pages checked (2026-11-15), and this page does not reproduce independent leaderboard figures. See How to Read LLM Benchmarks before comparing scores from other sources.

Deprecations and retirements

As of 2026-11-15, from OpenAI's deprecations page. A deprecated model stays callable until its shutdown date.

The same page lists mini models outside text generation, such as gpt-4o-mini-tts, gpt-realtime-mini and gpt-4o-mini-transcribe, with their own dates and audio or realtime replacements; those are out of scope here. The fine-tuned snapshot ft-o4-mini-2025-04-16 also shut down on 23 October 2026. Across providers, see the LLM Deprecation and Retirement Tracker.

Data handling and security

• Training use: OpenAI's data controls page states that, since 1 March 2023, data sent to the API is not used to train or improve models unless the customer opts in.
• Retention: the same page states that abuse-monitoring logs are kept for up to 30 days by default, longer where law requires.
• Zero Data Retention: available only to customers OpenAI approves; under it, content is excluded from abuse-monitoring logs and the store parameter is forced to false on /v1/responses and /v1/chat/completions.
• Keys and identities: small models are often called from many services at high volume, which multiplies API keys. See Workload Identity vs API Keys for LLM API Access, How to set spend limits on LLM API keys and the LLMjacking Guide.

How to choose a model: factual criteria

For small models from other providers with the same fields, see Small and Low-Cost LLMs Compared and the sibling page Gemini Flash and Flash-Lite Versions Compared.

Practitioner checklist

• Inventory every service that calls gpt-5-mini, gpt-5-nano, gpt-4o-mini, gpt-4.1-mini or gpt-5.4-nano, including pinned snapshot IDs.
• Put the shutdown dates (11 December 2026, 14 January 2027, 1 April 2027) on the change calendar with a test window before each.
• Run your evaluation set against gpt-6-luna or gpt-6.1-terra as named on the deprecations page, setting reasoning effort explicitly.
• Recalculate cost with the new input, cached input and output prices and the 272K long-prompt multiplier.
• Check whether any fine-tuned GPT-4o mini or GPT-4.1 mini model must be replaced by prompting or another route, since GPT-6 Luna lists fine-tuning as not supported.
• Confirm tool calls work on the target API: Chat Completions function calling on GPT-6 Luna needs reasoning_effort none.
• Review data residency, retention and Zero Data Retention needs before moving regulated workloads.
• Rotate and scope the API keys used by high-volume small-model callers, and set spend limits per key.

Standards and references

• OpenAI: Deprecations
• OpenAI: GPT-6 Luna model page
• OpenAI: GPT-6.1 Terra model page
• OpenAI: GPT-5.4 mini model page
• OpenAI: GPT-5.4 nano model page
• OpenAI: GPT-5 mini model page
• OpenAI: GPT-5 nano model page
• OpenAI: GPT-4.1 mini model page
• OpenAI: GPT-4o mini model page
• OpenAI: API pricing
• OpenAI: API changelog
• OpenAI: Data controls in the OpenAI platform

All sources accessed 2026-11-15.

Related NHI Mgmt Group resources: OpenAI Models, Small and Low-Cost LLMs Compared, LLM Deprecation and Retirement Tracker, Model Pricing Explained.