Prediction Held up AI assessment confidence: 85% certainty 3/5 debate potential 3/5

Altman: 10-million-token fast context windows are coming within months

Sam Altman · Building AGI in Real Time (OpenAI Dev Day 2024) · Oct 4, 2024 · at 2:05:30

OpenAI CEO Sam Altman answers an audience question regarding context window scaling at OpenAI DevDay 2024.

0:00 / 0:09exact quote · 9.5s

▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark

“Even getting to the, like, Ten million tokens of very fast and accurate context, which I expect to measure in, like, months, something like that.”

quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →

The claim, plainly

A fast and accurate AI model context window of at least 10 million tokens will be achieved within months.

a quote alone can be ambiguous, so this restates it with the references and context of the conversation around it resolved. AI-written; the tape below the player is the source.

Assessment

On October 4, 2024, then-OpenAI CEO Sam Altman stated he expected AI context windows of at least 10 million tokens with fast and accurate performance to be achieved within months. On April 5, 2025 (approximately six months later), Meta publicly released Llama 4 Scout under an open-weights license, officially specifying a 10-million-token context window alongside needle-in-a-haystack retrieval benchmarks demonstrating long-context recall. This fulfilled the prediction within the stated timeframe.

On April 5, 2025, Meta publicly released Llama 4 Scout with an officially supported 10-million-token context window and needle-in-a-haystack retrieval evaluations.
Google had previously demonstrated 10-million-token context retrieval experimentally in Gemini 1.5 research evaluations in February 2024.
Meta made Llama 4 Scout broadly available for download on Hugging Face and through third-party inference providers prior to the October 4, 2025 horizon.

Checked against

[1] arize.com [2] blog.google [3] wikipedia.org [4] hidekazu-konishi.com [5] google.com [6] ycombinator.com [7] lablab.ai [8] huggingface.co

10 million tokens context window OpenAI Google Anthropic 2025 Magic AI LTM 2 mini 100 million tokens August 2024 "10 million tokens" context window model 2024 2025 10M context window LLM benchmark needle in a haystack 2024 2025 10 million token context window AI model 2024 2025 "Llama 4 Scout" "10,000,000" OR "10M" AI model 10 million context window "2024" OR "2025" "Llama 4 Scout" Meta context window

assessed as of 2026-09-02

This assessment is an AI reading of a public statement, made against the cited references with the confidence shown. It can be wrong. The quote, the tape and the references are all on this page; if something is off, use report an error below. full disclaimer →

Mentioned

10 million tokens context window

⚑ Spot an error? Report it

More from Sam Altman

Assertion Not checkable as stated

Altman: OpenAI reached Level 2 AGI with o1

“I think we clearly got to level two, or we clearly got to level two with O-one.”

Sam Altman Oct 4, 2024 ▶ 1:24:50 Building AGI in Real Time (OpenAI Dev Day 2024)

Assertion Not checkable as stated

Altman: o1 is OpenAI's most aligned model ever by a lot

“And O-one is obviously our most capable model ever, but it's also our most aligned model ever by a lot.”

Sam Altman Oct 4, 2024 ▶ 1:33:35 Building AGI in Real Time (OpenAI Dev Day 2024)

Prediction Not checkable as stated

Altman: Infinite AI context windows will happen within a decade

“That obviously takes some research breakthroughs, but I assume that infinite context will happen at some point. At some point, it's like, less than a decade.”

Sam Altman Oct 4, 2024 ▶ 2:05:18 Building AGI in Real Time (OpenAI Dev Day 2024)

Prediction Not checkable as stated

Altman: AI will dynamically render custom real-time interfaces for any request

“At some point in not that many years in the future, you'll walk up to a piece of glass, you will say whatever you want they will have, like, there will be incredible reasoning models, agents connected to everything, there will be a video model streaming back t…”

Sam Altman Oct 4, 2024 ▶ 2:08:17 Building AGI in Real Time (OpenAI Dev Day 2024)

Altman: 10-million-token fast context windows are coming within months

More from Sam Altman

Turn any episode into a week of clips.