{"id":"de4d847c49932efff3e089902a959f0a568a0b6b54bd43bb795def1a0c56f005","title":"xAI announces Grok-2 and Grok-2 mini","argument":"xAI published its Grok-2 announcement on August 13, 2024. xAI presents Grok-2 as an early preview and Grok-2 mini as its smaller sibling. Both models are in beta on 𝕏, and xAI says enterprise API access is planned for later that month. xAI points to Chatbot Arena and academic benchmark results, reporting improvements over Grok-1.5 on the tests it evaluated. It also describes an internal evaluation in which AI Tutors compare two Grok responses, focusing on instruction following and factual accuracy. xAI says Grok-2 improved at reasoning with retrieved content and tool use. Its planned applications include enhanced search, deeper insights on 𝕏 posts, and improved replies. xAI says it is advancing core reasoning capabilities with a new compute cluster. The post reports evaluation results, but does not establish how they translate to reliability in everyday use.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"The post reports xAI's benchmark and internal evaluation results; it does not establish how those results translate to reliability in everyday use. The post says Grok-2 MMLU, MMLU-Pro, MMMU and MathVista were evaluated using 0-shot CoT, and notes that scores for several competing models came from earlier releases. Text as retrieved on October 24, 2026; originally published August 13, 2024.","speaker":{"name":"xAI","roleAtSource":"Publisher"},"source":{"title":"Grok-2 Beta Release","publisher":"xAI","format":"Article","url":"https://x.ai/news/grok-2","publishedAt":"2024-08-13T00:00:00Z","dateBasis":"Date stated on the publisher’s page."},"evidence":[{"text":"xAI announced Grok-2 and Grok-2 mini as beta models on 𝕏 and said both would become available through its enterprise API later that month.","excerpt":"Grok-2 and Grok-2 mini are currently in beta on 𝕏, and we are also making both models available through our enterprise API later this month.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI reports that Grok-2 and Grok-2 mini improved over Grok-1.5 on the academic benchmarks it evaluated.","excerpt":"Both Grok-2 and Grok-2 mini demonstrate significant improvements over our previous Grok-1.5 model.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says its internal evaluation used AI Tutors to compare two responses and assess instruction following and factual accuracy.","excerpt":"During each interaction, the AI Tutors are presented with two responses generated by Grok. They select the superior response based on specific criteria outlined in our guidelines.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says Grok-2 improved in reasoning with retrieved content and tool use, including identifying missing information and discarding irrelevant posts.","excerpt":"Grok-2 has shown significant improvements in reasoning with retrieved content and in its tool use capabilities, such as correctly identifying missing information, reasoning through sequences of events, and discarding irrelevant posts.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says its next steps include search, insights on 𝕏 posts, and improved reply functions, and says it is focusing on reasoning with a new compute cluster.","excerpt":"Our focus is on advancing core reasoning capabilities with our new compute cluster.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"Escape Meridian reads the benchmark results as evidence about performance on the selected tests, not as a direct measure of reliability in everyday use.","excerpt":"We evaluated the Grok-2 models across a series of academic benchmarks that included reasoning, reading comprehension, math, science, and coding.","url":"https://x.ai/news/grok-2","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://x.ai/news/grok-2","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well do Grok-2’s benchmark and internal evaluation results predict its reliability in everyday use?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"xAI announces Grok-2 and Grok-2 mini","summary":"xAI published its Grok-2 announcement on August 13, 2024. xAI presents Grok-2 as an early preview and Grok-2 mini as its smaller sibling. Both models are in beta on 𝕏, and xAI says enterprise API access is planned for later that month. xAI points to Chatbot Arena and academic benchmark results, reporting improvements over Grok-1.5 on the tests it evaluated. It also describes an internal evaluation in which AI Tutors compare two Grok responses, focusing on instruction following and factual accuracy. xAI says Grok-2 improved at reasoning with retrieved content and tool use. Its planned applications include enhanced search, deeper insights on 𝕏 posts, and improved replies. xAI says it is advancing core reasoning capabilities with a new compute cluster. The post reports evaluation results, but does not establish how they translate to reliability in everyday use.","question":"How well do Grok-2’s benchmark and internal evaluation results predict its reliability in everyday use?","claims":[{"text":"xAI announced Grok-2 and Grok-2 mini as beta models on 𝕏 and said both would become available through its enterprise API later that month.","excerpt":"Grok-2 and Grok-2 mini are currently in beta on 𝕏, and we are also making both models available through our enterprise API later this month.","support":"source-stated"},{"text":"xAI reports that Grok-2 and Grok-2 mini improved over Grok-1.5 on the academic benchmarks it evaluated.","excerpt":"Both Grok-2 and Grok-2 mini demonstrate significant improvements over our previous Grok-1.5 model.","support":"source-stated"},{"text":"xAI says its internal evaluation used AI Tutors to compare two responses and assess instruction following and factual accuracy.","excerpt":"During each interaction, the AI Tutors are presented with two responses generated by Grok. They select the superior response based on specific criteria outlined in our guidelines.","support":"source-stated"},{"text":"xAI says Grok-2 improved in reasoning with retrieved content and tool use, including identifying missing information and discarding irrelevant posts.","excerpt":"Grok-2 has shown significant improvements in reasoning with retrieved content and in its tool use capabilities, such as correctly identifying missing information, reasoning through sequences of events, and discarding irrelevant posts.","support":"source-stated"},{"text":"xAI says its next steps include search, insights on 𝕏 posts, and improved reply functions, and says it is focusing on reasoning with a new compute cluster.","excerpt":"Our focus is on advancing core reasoning capabilities with our new compute cluster.","support":"source-stated"},{"text":"Escape Meridian reads the benchmark results as evidence about performance on the selected tests, not as a direct measure of reliability in everyday use.","excerpt":"We evaluated the Grok-2 models across a series of academic benchmarks that included reasoning, reading comprehension, math, science, and coding.","support":"interpretation"}],"limitations":["The post reports xAI's benchmark and internal evaluation results; it does not establish how those results translate to reliability in everyday use.","The post says Grok-2 MMLU, MMLU-Pro, MMMU and MathVista were evaluated using 0-shot CoT, and notes that scores for several competing models came from earlier releases.","Text as retrieved on October 24, 2026; originally published August 13, 2024."],"mediaId":null,"correctionReason":null,"id":"de4d847c49932efff3e089902a959f0a568a0b6b54bd43bb795def1a0c56f005","publicationId":"93bce1c5-b3b1-f0f0-0efb-e4c95c9f4efa","version":1,"sourceUrl":"https://x.ai/news/grok-2","sourceTitle":"Grok-2 Beta Release","publisher":"xAI","publishedAt":"2024-08-13T00:00:00Z","originalPublished":{"value":"2024-08-13T00:00:00Z","precision":"instant","basis":"backfill:article_published"},"firstObservedAt":"2026-10-18T16:39:19.660Z","createdAt":"2026-11-17T17:19:05.435Z","historical":true,"entityIds":["xai"],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://x.ai/news/grok-2","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"e91c6a259befb5b0b0abb9292a69b70a719f5313594d117f0bfd4c1ea91dfae0","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-24T08:36:05.089Z"},"provenance":{"textHash":"ff3f929fc583b145382ec13015c3ae1728e2f25a229cb6080103683daa3735a8","observationId":"a2279f638c0291a619d7b725cbcb85bf6ef62b495017720a25c82e3613c662c0","retrievedAt":"2026-10-24T08:37:34.563Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"bd31dd25-c04e-d164-8bf1-e9202e94f1bc"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"bc2e3f3f-4ed4-fc06-862b-fe18c50e174b"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-17T17:14:27.432Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/xai-announces-grok-2-and-grok-2-mini-de4d847c","formats":{"html":"/readings/xai-announces-grok-2-and-grok-2-mini-de4d847c","markdown":"/readings/xai-announces-grok-2-and-grok-2-mini-de4d847c.md","json":"/readings/xai-announces-grok-2-and-grok-2-mini-de4d847c.json"},"attribution":"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.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-api","name":"API","explanation":"An application programming interface: a defined way for software to request data or actions from another system.","referenceUrl":"https://developer.mozilla.org/en-US/docs/Glossary/API"},{"id":"term-benchmark","name":"Benchmark","explanation":"A defined set of tests used to compare systems. Results depend on the tasks and measures selected.","referenceUrl":"https://arxiv.org/abs/2211.09110"},{"id":"term-tool-use","name":"Tool use","explanation":"A model's ability to operate other software, such as a search engine, a calculator or a code runner, while it works.","referenceUrl":"https://docs.anthropic.com/en/docs/build-with-claude/tool-use/overview"},{"id":"term-ai-compute","name":"AI compute","explanation":"The computing resources, meaning chips, data centers and electricity, used to train or run AI models.","referenceUrl":"https://developers.google.com/machine-learning/glossary#compute"},{"id":"term-chain-of-thought","name":"Chain of thought","explanation":"The step-by-step working a model writes out before giving its final answer.","referenceUrl":"https://developers.google.com/machine-learning/glossary#chain-of-thought-prompting"}],"entities":[{"id":"xai","name":"xAI","kind":"organization","context":"xAI develops the Grok family of AI models and publishes its own release claims and evaluations.","referenceUrls":["https://x.ai/","https://x.ai/news/grok-4-7"],"path":"/entities/xai","json":"/entities/xai.json","markdown":"/entities/xai.md"},{"id":"grok","name":"Grok","kind":"model","context":"Grok is a family of AI models from xAI, with capability claims and evaluations published by its developer.","referenceUrls":["https://x.ai/grok","https://x.ai/news/grok-4-7"],"path":"/entities/grok","json":"/entities/grok.json","markdown":"/entities/grok.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"de4d847c49932efff3e089902a959f0a568a0b6b54bd43bb795def1a0c56f005","revision":"93bce1c5-b3b1-f0f0-0efb-e4c95c9f4efa","revisionKind":"publication-id","title":"xAI announces Grok-2 and Grok-2 mini","url":"https://escapemeridian.com/readings/xai-announces-grok-2-and-grok-2-mini-de4d847c"},"source":{"type":"source","id":"https://x.ai/news/grok-2","revision":"ff3f929fc583b145382ec13015c3ae1728e2f25a229cb6080103683daa3735a8","revisionKind":"sha256-source-text","title":"Grok-2 Beta Release","url":"https://x.ai/news/grok-2"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"de4d847c49932efff3e089902a959f0a568a0b6b54bd43bb795def1a0c56f005","title":"xAI announces Grok-2 and Grok-2 mini","argument":"xAI published its Grok-2 announcement on August 13, 2024. xAI presents Grok-2 as an early preview and Grok-2 mini as its smaller sibling. Both models are in beta on 𝕏, and xAI says enterprise API access is planned for later that month. xAI points to Chatbot Arena and academic benchmark results, reporting improvements over Grok-1.5 on the tests it evaluated. It also describes an internal evaluation in which AI Tutors compare two Grok responses, focusing on instruction following and factual accuracy. xAI says Grok-2 improved at reasoning with retrieved content and tool use. Its planned applications include enhanced search, deeper insights on 𝕏 posts, and improved replies. xAI says it is advancing core reasoning capabilities with a new compute cluster. The post reports evaluation results, but does not establish how they translate to reliability in everyday use.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"The post reports xAI's benchmark and internal evaluation results; it does not establish how those results translate to reliability in everyday use. The post says Grok-2 MMLU, MMLU-Pro, MMMU and MathVista were evaluated using 0-shot CoT, and notes that scores for several competing models came from earlier releases. Text as retrieved on October 24, 2026; originally published August 13, 2024.","speaker":{"name":"xAI","roleAtSource":"Publisher"},"source":{"title":"Grok-2 Beta Release","publisher":"xAI","format":"Article","url":"https://x.ai/news/grok-2","publishedAt":"2024-08-13T00:00:00Z","dateBasis":"Date stated on the publisher’s page."},"evidence":[{"text":"xAI announced Grok-2 and Grok-2 mini as beta models on 𝕏 and said both would become available through its enterprise API later that month.","excerpt":"Grok-2 and Grok-2 mini are currently in beta on 𝕏, and we are also making both models available through our enterprise API later this month.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI reports that Grok-2 and Grok-2 mini improved over Grok-1.5 on the academic benchmarks it evaluated.","excerpt":"Both Grok-2 and Grok-2 mini demonstrate significant improvements over our previous Grok-1.5 model.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says its internal evaluation used AI Tutors to compare two responses and assess instruction following and factual accuracy.","excerpt":"During each interaction, the AI Tutors are presented with two responses generated by Grok. They select the superior response based on specific criteria outlined in our guidelines.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says Grok-2 improved in reasoning with retrieved content and tool use, including identifying missing information and discarding irrelevant posts.","excerpt":"Grok-2 has shown significant improvements in reasoning with retrieved content and in its tool use capabilities, such as correctly identifying missing information, reasoning through sequences of events, and discarding irrelevant posts.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"xAI says its next steps include search, insights on 𝕏 posts, and improved reply functions, and says it is focusing on reasoning with a new compute cluster.","excerpt":"Our focus is on advancing core reasoning capabilities with our new compute cluster.","url":"https://x.ai/news/grok-2","relationship":"Source-stated"},{"text":"Escape Meridian reads the benchmark results as evidence about performance on the selected tests, not as a direct measure of reliability in everyday use.","excerpt":"We evaluated the Grok-2 models across a series of academic benchmarks that included reasoning, reading comprehension, math, science, and coding.","url":"https://x.ai/news/grok-2","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://x.ai/news/grok-2","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well do Grok-2’s benchmark and internal evaluation results predict its reliability in everyday use?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"xAI announces Grok-2 and Grok-2 mini","summary":"xAI published its Grok-2 announcement on August 13, 2024. xAI presents Grok-2 as an early preview and Grok-2 mini as its smaller sibling. Both models are in beta on 𝕏, and xAI says enterprise API access is planned for later that month. xAI points to Chatbot Arena and academic benchmark results, reporting improvements over Grok-1.5 on the tests it evaluated. It also describes an internal evaluation in which AI Tutors compare two Grok responses, focusing on instruction following and factual accuracy. xAI says Grok-2 improved at reasoning with retrieved content and tool use. Its planned applications include enhanced search, deeper insights on 𝕏 posts, and improved replies. xAI says it is advancing core reasoning capabilities with a new compute cluster. The post reports evaluation results, but does not establish how they translate to reliability in everyday use.","question":"How well do Grok-2’s benchmark and internal evaluation results predict its reliability in everyday use?","claims":[{"text":"xAI announced Grok-2 and Grok-2 mini as beta models on 𝕏 and said both would become available through its enterprise API later that month.","excerpt":"Grok-2 and Grok-2 mini are currently in beta on 𝕏, and we are also making both models available through our enterprise API later this month.","support":"source-stated"},{"text":"xAI reports that Grok-2 and Grok-2 mini improved over Grok-1.5 on the academic benchmarks it evaluated.","excerpt":"Both Grok-2 and Grok-2 mini demonstrate significant improvements over our previous Grok-1.5 model.","support":"source-stated"},{"text":"xAI says its internal evaluation used AI Tutors to compare two responses and assess instruction following and factual accuracy.","excerpt":"During each interaction, the AI Tutors are presented with two responses generated by Grok. They select the superior response based on specific criteria outlined in our guidelines.","support":"source-stated"},{"text":"xAI says Grok-2 improved in reasoning with retrieved content and tool use, including identifying missing information and discarding irrelevant posts.","excerpt":"Grok-2 has shown significant improvements in reasoning with retrieved content and in its tool use capabilities, such as correctly identifying missing information, reasoning through sequences of events, and discarding irrelevant posts.","support":"source-stated"},{"text":"xAI says its next steps include search, insights on 𝕏 posts, and improved reply functions, and says it is focusing on reasoning with a new compute cluster.","excerpt":"Our focus is on advancing core reasoning capabilities with our new compute cluster.","support":"source-stated"},{"text":"Escape Meridian reads the benchmark results as evidence about performance on the selected tests, not as a direct measure of reliability in everyday use.","excerpt":"We evaluated the Grok-2 models across a series of academic benchmarks that included reasoning, reading comprehension, math, science, and coding.","support":"interpretation"}],"limitations":["The post reports xAI's benchmark and internal evaluation results; it does not establish how those results translate to reliability in everyday use.","The post says Grok-2 MMLU, MMLU-Pro, MMMU and MathVista were evaluated using 0-shot CoT, and notes that scores for several competing models came from earlier releases.","Text as retrieved on October 24, 2026; originally published August 13, 2024."],"mediaId":null,"correctionReason":null,"id":"de4d847c49932efff3e089902a959f0a568a0b6b54bd43bb795def1a0c56f005","publicationId":"93bce1c5-b3b1-f0f0-0efb-e4c95c9f4efa","version":1,"sourceUrl":"https://x.ai/news/grok-2","sourceTitle":"Grok-2 Beta Release","publisher":"xAI","publishedAt":"2024-08-13T00:00:00Z","originalPublished":{"value":"2024-08-13T00:00:00Z","precision":"instant","basis":"backfill:article_published"},"firstObservedAt":"2026-10-18T16:39:19.660Z","createdAt":"2026-11-17T17:19:05.435Z","historical":true,"entityIds":["xai"],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://x.ai/news/grok-2","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"e91c6a259befb5b0b0abb9292a69b70a719f5313594d117f0bfd4c1ea91dfae0","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-24T08:36:05.089Z"},"provenance":{"textHash":"ff3f929fc583b145382ec13015c3ae1728e2f25a229cb6080103683daa3735a8","observationId":"a2279f638c0291a619d7b725cbcb85bf6ef62b495017720a25c82e3613c662c0","retrievedAt":"2026-10-24T08:37:34.563Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"bd31dd25-c04e-d164-8bf1-e9202e94f1bc"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"bc2e3f3f-4ed4-fc06-862b-fe18c50e174b"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-17T17:14:27.432Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"93bce1c5-b3b1-f0f0-0efb-e4c95c9f4efa","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The page is a single xAI announcement on xAI's own site, and the reading attributes its claims to xAI throughout. The benchmark and capability claims stay xAI's assertions, and the one inference is labeled as Escape Meridian's interpretation.","excerpt":"We are excited to release an early preview of Grok-2, a significant step forward from our previous model Grok-1.5"},{"dimension":"summary","status":"uncertain","explanation":"The summary drops the post's headline claim: xAI says an early Grok-2, tested under the name sus-column-r, was outperforming Claude 3.5 Sonnet and GPT-4-Turbo on the LMSYS leaderboard. Instead it only says xAI 'points to Chatbot Arena' results. It also never explains what Chatbot Arena is, a public comparison of chatbots, so a general reader cannot tell what those results mean. Smallest repair: add one attributed sentence stating that comparative claim and briefly describing Chatbot Arena. A clause saying Grok is xAI's AI assistant would also help.","excerpt":"At the time of this blog post, it is outperforming both Claude 3.5 Sonnet and GPT-4-Turbo."},{"dimension":"date","status":"pass","explanation":"The reading's date, August 13, 2024, matches the UTC calendar day of evidence.originalPublished. The planned API release is reported as a plan ('later that month'), not as something that happened. There is no hindsight framing. This is a publisher page, so no archive disclosure is needed.","excerpt":"we are also making both models available through our enterprise API later this month."},{"dimension":"support","status":"pass","explanation":"Each claim matches its quoted source text. The note that competitor scores came from earlier releases reflects the footnotes, which date them to the May 2024 and June 2024 releases. The limitation that benchmarks do not establish everyday reliability is labeled as interpretation, and the question does not assume an answer.","excerpt":"* GPT-4-Turbo and GPT-4o scores are from the May 2024 release."},{"dimension":"media","status":"pass","explanation":"No media was selected, and no verified media candidates were supplied, so there is no media to check.","excerpt":null}],"createdAt":"2026-11-17T17:14:27.432Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}