Log in Sign up
Coding Assistants OpenAI Raises GPT-6.1 Sol's Default Context in Codex and Work to 500K Tokens — and Drops the Long-Context Fee for Subscribers Claude Anthropic Pulled Its Free Startup Perks 48 Hours After Launching Them — Applications Came in 21 Times Faster Than the Previous Five Months AI Agents A 1.7B Model Just Beat Its 6B and 8B Rivals: DexForce's UNITAS Treats Robot Manipulation as 3D Points, Not Pixels Business ASML Is Raising Prices 10% on Every EUV and DUV Spare Part Sold in South Korea — and Samsung and SK Hynix Have Already Agreed Business Infinigence AI Closes a RMB 2.3 Billion Series C - the Largest Single Round Yet Recorded in China's AI-Infrastructure Sector News Anthropic Quietly Signed a Cloud Deal With Ex-Bitcoin Miner IREN as Its Leased Compute Pipeline Reaches 15 Gigawatts Security PoeLLM Turns a Poem on GitHub Into a Command Server - and 3,400 Exposed AI Servers Into a Mining Botnet Business Nvidia's Off-Balance-Sheet Guarantees Jump 188% in a Quarter to $530 Billion - SemiAnalysis Sees a Path Past $1 Trillion Coding Assistants OpenAI Raises GPT-6.1 Sol's Default Context in Codex and Work to 500K Tokens — and Drops the Long-Context Fee for Subscribers Claude Anthropic Pulled Its Free Startup Perks 48 Hours After Launching Them — Applications Came in 21 Times Faster Than the Previous Five Months AI Agents A 1.7B Model Just Beat Its 6B and 8B Rivals: DexForce's UNITAS Treats Robot Manipulation as 3D Points, Not Pixels Business ASML Is Raising Prices 10% on Every EUV and DUV Spare Part Sold in South Korea — and Samsung and SK Hynix Have Already Agreed Business Infinigence AI Closes a RMB 2.3 Billion Series C - the Largest Single Round Yet Recorded in China's AI-Infrastructure Sector News Anthropic Quietly Signed a Cloud Deal With Ex-Bitcoin Miner IREN as Its Leased Compute Pipeline Reaches 15 Gigawatts Security PoeLLM Turns a Poem on GitHub Into a Command Server - and 3,400 Exposed AI Servers Into a Mining Botnet Business Nvidia's Off-Balance-Sheet Guarantees Jump 188% in a Quarter to $530 Billion - SemiAnalysis Sees a Path Past $1 Trillion

OpenAI Raises GPT-6.1 Sol's Default Context in Codex and Work to 500K Tokens — and Drops the Long-Context Fee for Subscribers

Coding Assistants · By: AIReport Admin · 1 hour ago · 👁 6
OpenAI Raises GPT-6.1 Sol's Default Context in Codex and Work to 500K Tokens — and Drops the Long-Context Fee for Subscribers

OpenAI has lifted the default context length for GPT-6.1 Sol in Codex and Work from 272,000 to 500,000 tokens, roughly an 80% increase, according to OpenAI engineer Thibault Sottiaux. The change is meant to cut how often long sessions trigger compaction. OpenAI says subscription usage carries no long-context surcharge, though API users still pay higher rates above 272,000 tokens.

OpenAI has raised the default context window for GPT-6.1 Sol in Codex and ChatGPT Work from 272,000 tokens to 500,000 tokens, an increase of roughly 80%, according to an update posted on X by OpenAI's Thibault Sottiaux. The setting is a default, not a ceiling: users can still dial it up or down, and GPT-6.1 Sol's maximum stays at 1 million tokens.

The point of the change is compaction. When a long Codex session fills its context, the tool summarizes older parts of the history to free room, which keeps the session alive but can quietly drop detail. Sottiaux said the larger default "helps reduce occurrence of compactions" and helps with tasks "that require keeping a lot of context and nuance over periods of time," calling 500,000 tokens the new "sweet spot for most of our users" after testing. No test results were published, so the size of the practical improvement has not been independently checked.

The old default was not arbitrary. On OpenAI's API pricing, 272,000 input tokens is exactly the threshold where GPT-6.1 Sol switches to long-context rates: above that line a request is billed at $4 instead of $2 per million input tokens and $15 instead of $10 per million output tokens. The previous Codex default sat precisely below the point where costs jump.

That makes the second half of the announcement the more consequential one. Sottiaux said OpenAI is "not charging long context fees for GPT-6.1 Sol when drawing usage from the subscription." For paying ChatGPT users, the bigger default carries no surcharge. Developers paying through the API still pay long-context rates once a prompt passes 272,000 tokens, so the relief is confined to plan-based usage.

Two questions the post does not answer are whether longer sessions consume plan limits faster, since a larger context means more tokens are processed on every turn, and whether the new default reaches every paid tier or is still rolling out. OpenAI has not published a help center note or changelog entry on the change.

The update is labeled "Day 7" in a run of near-daily posts Sottiaux has been sharing about Codex and ChatGPT, a cadence that reads as a public commitment to ship visible improvements continuously rather than in occasional large releases. Earlier entries covered faster steering, a re-shipped Codex cloud environment and a reset of usage limits. It follows GPT-6.1 Sol's arrival as the main model in ChatGPT Work and Codex.

Context windows have become one of the least glamorous and most practically decisive numbers in the agent era. Long-running coding and research agents fail not because the model is not smart enough but because the history gets compressed and details are lost. Raising a default and removing the fee on it is a cheaper way to buy reliability than a new model generation, and it is aimed squarely at the workflows where agents run for hours rather than minutes.

The competitive signal is that the long-context premium is being absorbed rather than charged. Rival labs have spent the past month shipping million-token contexts and cutting cache pricing; OpenAI's answer here is not a bigger window but a cheaper one to actually use. If the pattern holds, extended context stops being a paid upgrade and becomes table stakes.

Comments (0)

Log in to join the discussion

Log In

No comments yet