# Anthropic describes extended thinking’s benefits and limits [Canonical reading](https://escapemeridian.com/readings/anthropic-describes-extended-thinking-s-benefits-and-limits-e1760e6a) Anthropic’s page, published February 24, 2025, describes extended thinking in Claude 3.7 Sonnet: users can turn it on or off, and developers can set a thinking budget. Anthropic says the same model spends more time and effort on an answer, with its thought process visible in raw form. The company presents visible thoughts as a way to inspect answers and study model behavior, but warns that they may include incorrect or misleading reasoning and may not faithfully represent why the model behaves as it does. Anthropic says monitoring current models’ thinking cannot support strong arguments about their safety. That gap matters because visible reasoning may look informative without revealing the factors behind a decision. Anthropic reports improved results on evaluations and describes the model using computer controls and playing Pokémon Red. It also reports gains from parallel test-time compute on GPQA, while noting that this approach was not available in the deployed model. These evaluation results do not establish everyday reliability. The post describes safeguards for visible thoughts and computer use, calls visible extended thinking a research preview, and says Anthropic will weigh its trade-offs for future releases. How well can visible thoughts help assess a model’s behavior if they may not represent the factors behind its decisions? 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 Anthropic · Article Original source date: Feb 24, 2025 (2025-02-24T00:00:00.000Z). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-15T06:03:23.419Z. First observed: 2026-10-20T11:57:19.987Z. Retrieved: 2026-10-29T18:07:44.868Z. - [Claude's extended thinking](https://www.anthropic.com/news/visible-extended-thinking) ## Limits Text as retrieved on October 29, 2026; originally published February 24, 2025. The page was edited after publication, so the retrieved text may include later additions; this reading does not treat later additions as part of the original publication. Anthropic’s reported evaluation results do not establish performance or reliability in everyday use. No verified media candidates were supplied. ## Open question How well can visible thoughts help assess a model’s behavior if, as Anthropic says, they may not represent the factors behind its decisions? ## Source-stated Anthropic says Claude 3.7 Sonnet’s extended thinking mode can be toggled by users, while developers can set a thinking budget. > users can toggle “extended thinking mode” on or off, directing the model to think more deeply about trickier questions1. And developers can even set a “thinking budget” - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says the visible thought process may contain incorrect or misleading reasoning and may not faithfully represent why the model behaves as it does. > we don’t know for certain that what’s in the thought process truly represents what’s going on in the model’s mind - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says monitoring current models’ thinking cannot support strong arguments about their safety. > This means we can’t rely on monitoring current models’ thinking to make strong arguments about their safety2. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic reports that parallel test-time compute scaling produced a GPQA score of 84.8% under the described experimental setup. > Using the equivalent compute of 256 independent samples, a learned scoring model, and a maximum 64k-token thinking budget, Claude 3.7 Sonnet achieved a GPQA score of 84.8% - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says its computer-use safeguards prevented prompt injection attacks 88% of the time, compared with 74% without the mitigations. > we now prevent these attacks 88% of the time4, up from 74% of the time without the mitigations. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic calls visible extended thinking in Claude 3.7 Sonnet a research preview. > the visible thought process in Claude 3.7 Sonnet should be considered a research preview. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source and contributor context Anthropic · 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. ## Test-time compute Extra computation a model spends while answering, for example by reasoning longer, rather than during training. - [Definition reference](https://arxiv.org/abs/2408.03314) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: 9e65f45c-4135-5975-3ec4-236944acb76e. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T05:59:05.789Z. 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: e1ebd29a154a8a2a01d5cb2bb63349146f61b1b377881f910fd68be5fda0fd7e. Source text hash: 10d04d381ebbec2f0a68e7f1a5d2b96551c757b7db875271c7e8dfcdae792791. ## Revision and source-check history ### Version 1 — 2026-11-15T06:03:23.419Z Anthropic’s page, published February 24, 2025, describes extended thinking in Claude 3.7 Sonnet: users can turn it on or off, and developers can set a thinking budget. Anthropic says the same model spends more time and effort on an answer, with its thought process visible in raw form. The company presents visible thoughts as a way to inspect answers and study model behavior, but warns that they may include incorrect or misleading reasoning and may not faithfully represent why the model behaves as it does. Anthropic says monitoring current models’ thinking cannot support strong arguments about their safety. That gap matters because visible reasoning may look informative without revealing the factors behind a decision. Anthropic reports improved results on evaluations and describes the model using computer controls and playing Pokémon Red. It also reports gains from parallel test-time compute on GPQA, while noting that this approach was not available in the deployed model. These evaluation results do not establish everyday reliability. The post describes safeguards for visible thoughts and computer use, calls visible extended thinking a research preview, and says Anthropic will weigh its trade-offs for future releases. How well can visible thoughts help assess a model’s behavior if they may not represent the factors behind its decisions? ## Original source Anthropic · Article Original source date: Feb 24, 2025 (2025-02-24T00:00:00.000Z). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-15T06:03:23.419Z. First observed: 2026-10-20T11:57:19.987Z. Retrieved: 2026-10-29T18:07:44.868Z. - [Claude's extended thinking](https://www.anthropic.com/news/visible-extended-thinking) ## Limits Text as retrieved on October 29, 2026; originally published February 24, 2025. The page was edited after publication, so the retrieved text may include later additions; this reading does not treat later additions as part of the original publication. Anthropic’s reported evaluation results do not establish performance or reliability in everyday use. No verified media candidates were supplied. ## Open question How well can visible thoughts help assess a model’s behavior if, as Anthropic says, they may not represent the factors behind its decisions? ## Source-stated Anthropic says Claude 3.7 Sonnet’s extended thinking mode can be toggled by users, while developers can set a thinking budget. > users can toggle “extended thinking mode” on or off, directing the model to think more deeply about trickier questions1. And developers can even set a “thinking budget” - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says the visible thought process may contain incorrect or misleading reasoning and may not faithfully represent why the model behaves as it does. > we don’t know for certain that what’s in the thought process truly represents what’s going on in the model’s mind - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says monitoring current models’ thinking cannot support strong arguments about their safety. > This means we can’t rely on monitoring current models’ thinking to make strong arguments about their safety2. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic reports that parallel test-time compute scaling produced a GPQA score of 84.8% under the described experimental setup. > Using the equivalent compute of 256 independent samples, a learned scoring model, and a maximum 64k-token thinking budget, Claude 3.7 Sonnet achieved a GPQA score of 84.8% - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic says its computer-use safeguards prevented prompt injection attacks 88% of the time, compared with 74% without the mitigations. > we now prevent these attacks 88% of the time4, up from 74% of the time without the mitigations. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source-stated Anthropic calls visible extended thinking in Claude 3.7 Sonnet a research preview. > the visible thought process in Claude 3.7 Sonnet should be considered a research preview. - [Supporting source](https://www.anthropic.com/news/visible-extended-thinking) ## Source and contributor context Anthropic · 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. ## Test-time compute Extra computation a model spends while answering, for example by reasoning longer, rather than during training. - [Definition reference](https://arxiv.org/abs/2408.03314) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## TIME TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services. - [Explore this context](/entities/time) - [Reference](https://time.com/) ## Editorial provenance Version 1. Publication ID: 9e65f45c-4135-5975-3ec4-236944acb76e. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T05:59:05.789Z. 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: e1ebd29a154a8a2a01d5cb2bb63349146f61b1b377881f910fd68be5fda0fd7e. Source text hash: 10d04d381ebbec2f0a68e7f1a5d2b96551c757b7db875271c7e8dfcdae792791. ### passed — 2026-11-15T05:59:05.789Z — anthropic/claude-opus-5.5 Publication: 9e65f45c-4135-5975-3ec4-236944acb76e attribution — pass: The reading consistently attributes the descriptions, warnings, benchmark results and safeguard figures to Anthropic's own post about Claude 3.7 Sonnet. The page is a single first-party article, not a listing. > With the new Claude 3.7 Sonnet, users can toggle “extended thinking mode” on or off summary — pass: The reading leads with the actor and the change: a toggleable extended thinking mode with a developer-set thinking budget and a visible raw thought process. It conveys the post's central tension, that visible reasoning helps trust and research but may not be faithful, and it reports Anthropic's research-preview framing. Test-time compute is defined beside the reading. GPQA is not explained but is not central. The question follows from the source's faithfulness concern. Omitting the ASL-2 determination does not distort the reading's focus. > Another issue is what’s known as “faithfulness”—we don’t know for certain that what’s in the thought process truly represents what’s going on in the model’s mind date — pass: The stated date, February 24, 2025, matches the UTC day of originalPublished. The page was edited after publication, but the cited passages concern the Claude 3.7 Sonnet launch and are plainly part of the original piece. The limitations disclose the later retrieval and the edit. No hindsight framing appears, and the plans for future releases remain attributed intentions. > We’ll weigh the pros and cons of revealing the thought process for future releases3. support — pass: Each claim matches the source text: the toggle and thinking budget, the incorrect or misleading thoughts, faithfulness, the safety-monitoring limit, GPQA 84.8% under the stated setup, prompt injection prevention of 88% versus 74%, and the research preview. The line "That gap matters" stays close to the source's own point that decisions rest on undiscussed factors. The everyday-reliability caveat is appropriately cautious. > Thus far, our results suggest that models very often make decisions based on factors that they don’t explicitly discuss in their thinking process. media — pass: No media was selected, and no verified media candidates were supplied, so no invalid or unverified media is attached. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/e1760e6a4ebca25e1cc44f0c1e49daa8373d60a6b7fd730f98fe408404774ef5/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/anthropic-describes-extended-thinking-s-benefits-and-limits-e1760e6a.json)