Pull to refresh
Logo
What Changed
Following Why Sign Up
OpenAI expands GPT-5 context window to one million tokens

OpenAI expands GPT-5 context window to one million tokens

New Capabilities

Latest GPT-5 generation models process more than 100 times more text per request than GPT-4, enabling single-request document and codebase generation

October 20th, 2026: 12-hour prompt caching added to the Responses API

Overview

Updated Oct 26

OpenAI's GPT-5 generation models, GPT-5.4 and GPT-5.5, accept up to one million tokens per request. That is enough for a full codebase or an entire book, and about 125 times GPT-4's 8,000-token window. The base GPT-5 API model takes 272,000 input tokens with a 400,000-token total context.

Pricing for the long-context models is public. GPT-5.5 costs $5 per million input tokens and $30 per million output. In October, OpenAI moved to GPT-6.1 Sol, which has a one-million-token maximum. On Oct. 11 its default context in Codex and ChatGPT Work rose from 272,000 to 500,000 tokens. On Oct. 20 it added 12-hour prompt caching, where cache writes cost 2.5 times the input price. API requests above 272,000 tokens still pay higher long-context rates.

Why it matters

A million-token context lets a single GPT-5 call digest a full codebase or book, collapsing multi-step AI pipelines into one request.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

1,000,000
GPT-5 context window (tokens)
Maximum tokens the top GPT-5 generation models can process per request.
125x
Context increase over GPT-4
Million-token window versus GPT-4's 8,000-token limit.
33%
Fewer false claims in GPT-5.4 vs GPT-5.2
On prompts users flagged for factual errors, per OpenAI's release notes.
$5
GPT-5.5 API price per 1M input tokens
Million-token model at $5 per million input; output is $30 per million. GPT-5.5 Pro is $30 input and $180 output.
82.7%
GPT-5.5 Terminal-Bench 2.0 score
Coding benchmark result OpenAI reported for GPT-5.5.

Voices

Curated perspectives — historical figures and your fellow readers.

Sign up to generate historical perspectives on this story.

Log in Sign Up

People Involved

Sam Altman
Sam Altman
Chief executive of OpenAI
Leads OpenAI through the GPT-5 release cycle
 Mira Murati
Mira Murati
Chief technology officer of OpenAI
Public face of GPT-5's technical capabilities

Organizations Involved

OpenAI
OpenAI
Artificial intelligence company
Developer of GPT-5 and the GPT-5-series models

The company behind GPT, ChatGPT, and the GPT-5-series models.

Timeline

March 2023 October 2026

8 events Latest: October 20th, 2026 · 3 weeks ago
Tap a bar to jump to that date

  1. 12-hour prompt caching added to the Responses API

    Latest Release

    Developers can keep cached prompts for 12 hours on GPT-5.6 and later models. A 12-hour cache write costs 2.5 times the input price, versus 1.25 times for the 30-minute default.

  2. GPT-6.1 Sol default context raised to 500,000 tokens in Codex and Work

    Release

    OpenAI raised the default context window for GPT-6.1 Sol in Codex and ChatGPT Work from 272,000 to 500,000 tokens. Its maximum stays at one million tokens.

  3. OpenAI unveils GPT-5 with million-token context

    Announcement

    OpenAI unveils GPT-5 with a one-million-token context, more than 100 times GPT-4's.

  4. GPT-5.5 reaches the API with pricing

    Release

    GPT-5.5 and GPT-5.5 Pro reach the API with a one-million-token window. GPT-5.5 is priced at $5 per million input tokens and $30 per million output; GPT-5.5 Pro is $30 and $180.

  5. GPT-5.4 ships with million-token context

    Release

    GPT-5.4 ships with a one-million-token context window and native computer-use in Codex. OpenAI offered the long context experimentally in Codex at double the standard usage rate.

  6. GPT-5 launches as OpenAI's flagship

    Launch

    GPT-5 launches as OpenAI's flagship with built-in reasoning, available to all users.

  7. GPT-5 ships to the API with published pricing

    Release

    OpenAI priced GPT-5 at $1.25 per million input tokens and $10 per million output, with mini at $0.25 and nano at $0.05 input. The API models accept 272,000 input tokens with a 400,000 total context. GPT-5 scored 74.9% on SWE-bench Verified.

  8. GPT-4 launches with 8,000-token context

    Launch

    GPT-4 launches with an 8,000-token context window, the baseline for GPT-5.

Scenarios

1

Million-token context reaches all GPT-5 API tiers

Likely Resolves by Q2 2027

Discussed by: OpenAI developer documentation and Codex release notes

OpenAI has given GPT-5.4 experimental one-million-token support in Codex at double the standard usage rate, and GPT-5.5 a one-million-token API window. If developer demand holds, the full context opens to standard non-Pro tiers and appears in published API documentation.

2

Rivals undercut OpenAI on long-context pricing

Possible Resolves by Q2 2027

Discussed by: The Decoder, industry pricing trackers

Google and Anthropic already ship million-token models. OpenAI prices million-token usage at twice the standard rate in Codex and lists GPT-5.5 Pro at $30 per million input and $180 per million output tokens. If competitors undercut those rates, long-context work flows to them; if they match, a pricing contest follows.

3

Independent tests find GPT-5 accuracy slips at long context

Possible Resolves by Q2 2027

Discussed by: Third-party evaluation labs such as Artificial Analysis

OpenAI says GPT-5.4's claims are 33% less likely to be false than GPT-5.2's on flagged prompts. Whether accuracy holds at a million tokens, retrieving facts across a full codebase, is a separate question that independent evaluations will test as long-context use spreads.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

February 2024

Gemini 1.5 Pro (2024)

Google gave developers a one-million-token context window in its Gemini 1.5 Pro model, letting it process an entire book's worth of text or hours of video in one pass.

Then

Google demonstrated that million-token contexts were feasible and marketed Gemini's capacity against OpenAI's smaller windows.

Now

The one-million-token bar became the industry comparison point for long-context models.

Why this matters now

OpenAI's million-token window matches a threshold Google set years earlier, rather than setting a new one.

March 2023

GPT-4 launch (2023)

OpenAI released GPT-4 with an 8,000-token context window, up from roughly 4,000 in earlier GPT-3.5-era models.

Then

The larger window let developers build applications that could process longer documents and hold more instructions.

Now

Each context jump has unlocked new application categories, from document analysis to multi-step agents.

Why this matters now

GPT-4's 8,000-token limit is the baseline OpenAI now measures against; the million-token window is a 125-fold jump.

Sources

(14)
+8
UXC News OpenAI The Decoder AI In Context DiffVibe OpenAI OpenAI DailyCited AskGPU AIReport.NET OpenAI OpenAI OpenAI OpenAI