MIT Technology Review · Jan 31, 2025
MIT Technology Review reports OpenAI bringing o3-mini to free ChatGPT users
Markdown · JSON · Original source
MIT Technology Review reported on January 31, 2025, that was releasing its o3-mini to people using free ChatGPT. The article describes reasoning models as working through problems step by step, which can make answers more thorough but also mean longer waits. OpenAI claimed o3-mini responds faster than o1-mini. The article presents complex problem-solving as a possible use, alongside a cost trade-off: o3-mini costs less than o1-mini per input token but remains more expensive to run than GPT-4o mini. On safety, the article says OpenAI’s o3-mini is better than non-reasoning models at and challenging safety evaluations. OpenAI rated it “medium risk” on model autonomy because of its coding-task performance and potential for self-improvement and AI research acceleration. The article says the model remains bad at real-world research; if it were better, OpenAI would rate it high risk and restrict its release. How these capabilities and safety findings hold up in real-world use by free ChatGPT users remains unknown.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
MIT Technology Review · Article
Source details
Original source date: Jan 31, 2025 (2025-01-31).
Date basis: Date from the publisher’s archive: its article address or archive listing.
Added to Escape Meridian: 2026-11-16T03:23:23.448Z. First observed: 2026-10-24T01:46:17.457Z. Retrieved: 2026-10-24T01:47:35.204Z.
Referenced by
Recorded citations in public readings. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published January 31, 2025. The article gives no measurement or method for its statement that are estimated to have higher energy costs.
Open question
How do o3-mini’s reasoning capabilities and safety findings hold up in real-world use by free ChatGPT users?
Source-stated
MIT Technology Review reports that OpenAI made o3-mini available to people using the free version of ChatGPT.
OpenAI is releasing a new reasoning model, o3-mini, to people who use the free version of ChatGPT.
Source-stated
The article says reasoning models use chain of thought to work through problems step by step, and notes this can mean longer waits for answers.
Reasoning models use a “chain of thought” technique to generate responses, essentially working through a problem presented to the model step by step.
Source-stated
OpenAI claims o3-mini responds faster than o1-mini.
OpenAI claims that o3-mini responds 24% faster than o1-mini.
Source-stated
The article says o3-mini costs less than o1-mini per input token but remains more expensive to run than GPT-4o mini.
The company says its new model, o3-mini, costs 63% less than o1-mini per input token However, at $1.10 per million input tokens, it is still about seven times more expensive to run than GPT-4o mini.
Source-stated
The article says o3-mini is better than non-reasoning models at jailbreaking and challenging safety evaluations, and that OpenAI rated it “medium risk” on model autonomy.
But the company has found that o3-mini, like the o1 model, is significantly better than non-reasoning models at jailbreaking and “challenging safety evaluations”—essentially, it’s much harder to control a reasoning model given its advanced capabilities.
Source-stated
The article says OpenAI did not rate o3-mini high risk or restrict its release because it remains bad at real-world research.
That said, the model is still bad at real-world research. If it were better at that, it would be rated as high risk, and OpenAI would restrict the model’s release.
Source and contributor context
MIT Technology Review · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Reasoning model
A language model trained to work through a problem in written steps before it answers.
Jailbreak
A prompt written to trick an AI system into ignoring its safety rules.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Editorial provenance
Version 1. Publication ID: a1984bde-4aaa-cf87-8efc-33c4aa2aab2b.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T03:19:17.208Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: d142ab8f268af2678fd41fe53d52f711dd353b8b443564bd9a2ce7306413c35d. Source text hash: fd419cc0d994bec2c7f3597c063bd2c1ea3d0d91dbe6b068b4ea0c608710d2b4.
Revision and source-check history
Version 1 · 2026-11-16T03:23:23.448Z
MIT Technology Review reports OpenAI bringing o3-mini to free ChatGPT users
MIT Technology Review reported on January 31, 2025, that was releasing its o3-mini to people using free ChatGPT. The article describes reasoning models as working through problems step by step, which can make answers more thorough but also mean longer waits. OpenAI claimed o3-mini responds faster than o1-mini. The article presents complex problem-solving as a possible use, alongside a cost trade-off: o3-mini costs less than o1-mini per input token but remains more expensive to run than GPT-4o mini. On safety, the article says OpenAI’s o3-mini is better than non-reasoning models at and challenging safety evaluations. OpenAI rated it “medium risk” on model autonomy because of its coding-task performance and potential for self-improvement and AI research acceleration. The article says the model remains bad at real-world research; if it were better, OpenAI would rate it high risk and restrict its release. How these capabilities and safety findings hold up in real-world use by free ChatGPT users remains unknown.
Original source
MIT Technology Review · Article
Source details
Original source date: Jan 31, 2025 (2025-01-31).
Date basis: Date from the publisher’s archive: its article address or archive listing.
Added to Escape Meridian: 2026-11-16T03:23:23.448Z. First observed: 2026-10-24T01:46:17.457Z. Retrieved: 2026-10-24T01:47:35.204Z.
Limits
Text as retrieved on October 24, 2026; originally published January 31, 2025. The article gives no measurement or method for its statement that are estimated to have higher energy costs.
Open question
How do o3-mini’s reasoning capabilities and safety findings hold up in real-world use by free ChatGPT users?
Source-stated
MIT Technology Review reports that OpenAI made o3-mini available to people using the free version of ChatGPT.
OpenAI is releasing a new reasoning model, o3-mini, to people who use the free version of ChatGPT.
Source-stated
The article says reasoning models use chain of thought to work through problems step by step, and notes this can mean longer waits for answers.
Reasoning models use a “chain of thought” technique to generate responses, essentially working through a problem presented to the model step by step.
Source-stated
OpenAI claims o3-mini responds faster than o1-mini.
OpenAI claims that o3-mini responds 24% faster than o1-mini.
Source-stated
The article says o3-mini costs less than o1-mini per input token but remains more expensive to run than GPT-4o mini.
The company says its new model, o3-mini, costs 63% less than o1-mini per input token However, at $1.10 per million input tokens, it is still about seven times more expensive to run than GPT-4o mini.
Source-stated
The article says o3-mini is better than non-reasoning models at jailbreaking and challenging safety evaluations, and that OpenAI rated it “medium risk” on model autonomy.
But the company has found that o3-mini, like the o1 model, is significantly better than non-reasoning models at jailbreaking and “challenging safety evaluations”—essentially, it’s much harder to control a reasoning model given its advanced capabilities.
Source-stated
The article says OpenAI did not rate o3-mini high risk or restrict its release because it remains bad at real-world research.
That said, the model is still bad at real-world research. If it were better at that, it would be rated as high risk, and OpenAI would restrict the model’s release.
Source and contributor context
MIT Technology Review · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Reasoning model
A language model trained to work through a problem in written steps before it answers.
Jailbreak
A prompt written to trick an AI system into ignoring its safety rules.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Editorial provenance
Version 1. Publication ID: a1984bde-4aaa-cf87-8efc-33c4aa2aab2b.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T03:19:17.208Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: d142ab8f268af2678fd41fe53d52f711dd353b8b443564bd9a2ce7306413c35d. Source text hash: fd419cc0d994bec2c7f3597c063bd2c1ea3d0d91dbe6b068b4ea0c608710d2b4.
passed · 2026-11-16T03:19:17.208Z · anthropic/claude-opus-5.5
Publication: a1984bde-4aaa-cf87-8efc-33c4aa2aab2b
attribution — pass: The reading consistently attributes its report to MIT Technology Review and attributes the speed, cost, risk-rating and safety statements to OpenAI or to the article. The source is a single news article. The related-story list at the bottom is page furniture and is not used by the reading.
OpenAI is releasing a new reasoning model, o3-mini, to people who use the free version of ChatGPT.
summary — pass: The reading leads with the actor (OpenAI) and the change (o3-mini reaching free ChatGPT users). It explains reasoning models in plain terms, and Jailbreak and Reasoning model are defined beside the reading. It also covers the article's cost trade-off and safety findings. The jailbreaking wording mirrors the source's own phrasing. Omitting the DeepSeek comparison and deliberative alignment does not distort the article's meaning. The question is open, not leading.
This will mark the first time that the vast majority of people will have access to one of OpenAI’s reasoning models
date — pass: The reading's January 31, 2025 date matches originalPublished at day precision. The source's 'On Thursday' Microsoft reference fits that dateline. Sidebar items describing later events, such as the AlphaGo ten-year piece, are possible later page additions; the reading neither uses nor relies on them. The page is a publisher page, so no archive disclosure is needed, and modifiedAt falls on the same day.
Ten years after AlphaGo’s match against Go champion Lee Sedol
support — pass: Each claim is supported by its excerpt. Claim 6 reasonably restates the source's conditional statement that a better model would be rated high risk and restricted. The energy-cost limitation accurately notes the article's unsourced estimate. The concluding statement about unknown real-world performance is framed as an open question, not a claim.
That said, the model is still bad at real-world research. If it were better at that, it would be rated as high risk, and OpenAI would restrict the model’s release.
media — pass: No media candidates were supplied, and the reading selects none, which is appropriate.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIA language model trained to work through a problem in written steps before it answers.
Definition referenceA prompt written to trick an AI system into ignoring its safety rules.
Definition referenceA language model trained to work through a problem in written steps before it answers.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIA language model trained to work through a problem in written steps before it answers.
Definition referenceA prompt written to trick an AI system into ignoring its safety rules.
Definition referenceA language model trained to work through a problem in written steps before it answers.
Definition reference