{"success":true,"data":{"pressRelease":{"id":"49406","rtpr_id":"nBw1rVK10a","ticker":"TEM","exchange":"NASDAQ","all_tickers":["TEM"],"title":"Tempus Announces Initial Results from its Multimodal Foundation Model Efforts for Novel and Scalable Insight Generation in Oncology","author":"Business Wire","published_at":"2026-05-29T13:15:00.172Z","article_body":"Tempus Announces Initial Results from its Multimodal Foundation Model Efforts\nfor Novel and Scalable Insight Generation in Oncology\n\nTempus AI, Inc. (NASDAQ: TEM), a technology company leading the adoption of AI\nto advance precision medicine, today announced the latest results from its\nmission to build Multimodal Foundation Models at the 2026 American Society of\nClinical Oncology (ASCO) Annual Meeting. Tempus has been building novel\nbiological foundation models and agentic workflows by leveraging its more than\n500 petabytes of rich, molecularly grounded data—more than 45 million total\nde-identified patient journeys, 1.5 million with sequenced data, and over\n400,000 records in cancer with full genomic, transcriptomic, imaging, and\nclinical data. These efforts transform this data into unified patient\nrepresentations, unlocking actionable insights to speed precision medicine\nefforts in both the clinic and drug development.\n\nTempus’ latest multimodal, transformer-based model was trained on 2.5\nmillion longitudinal records encapsulating more than 250 million pages of\nclinical notes, 450,000 digitized medical images, and 500,000 genomic and\ntranscriptomic sequences. By aggregating modalities derived from\nbillion-parameter foundation models, it is designed to address thousands of\nprediction objectives anchored in overall survival (OS) and progression-free\nsurvival (PFS), without requiring additional data or model fine-tuning. This\narchitecture significantly reduces the time and data required to produce\nhundreds of clinically relevant insights for clinical trial design, patient\nrisk prediction, and novel multimodal diagnostics.\n\nIn a zero-shot setting without any further training, Tempus leveraged this\npatient trajectory model and a series of agentic workflows to unlock insights\nin a cohort of more than 1.2 million de-identified records with robust\nmultimodal data.\n\nAs a primary proof of concept, Tempus’ model was utilized to analyze\nEGFR-mutant NSCLC patients treated with osimertinib, the current frontline\nstandard of care third-generation EGFR-inhibitor. Tempus assessed whether the\nmodel could accurately stratify response to the standard of care treatment of\nosimertinib in patients with known clinically actionable biomarkers like EGFR.\nThis tests the model’s ability to learn a composite of features that predict\nresponse outside of known biological and clinical features to identify\npatients more likely to experience poor response to current therapies.\n\nWithout any pre-specified training, the model demonstrated:\n\n\n * Predictive Accuracy: A C-index of 0.802 for overall survival (p value <\n0.001).\n\n * Significant Survival Stratification: Hazard Ratio of 4.536 (95% CI: [3.289,\n6.255]) between high- and low-risk subgroups.\n\nThe model was evaluated against a range of mutation profiles (including more\nthan 30 EGFR-specific features including L858R, Exon 19 del); and the results\nstill produced prognostic value independent of molecular and clinical\nsubgroups, significantly stratifying the overall survival of TP53(+) patients\n(HR=5.96), as well as progression-free survival in patients without CNS\nmetastasis (HR=1.94). This outcome is just one example of how Tempus’ model\ncan produce novel clinical insights, as similar results have been observed for\npredicting overall survival in other patient populations.\n\nTempus’ multimodal patient trajectory models have also demonstrated\nsubstantial utility for drug development activities. For example, the models\nhave successfully predicted outcomes of patient cohorts that mirrored\npre-established clinical trials. In NSCLC, Tempus was able to assess cohorts\nmirroring those of practice-changing clinical trials (KEYNOTE-189, FLAURA-2,\nand DESTINY), and the multimodal patient trajectory model was able to\nzero-shot outperform standard approaches, such as Cox-PH modeling. These\noutcomes indicate substantial utility for biopharmaceutical developers,\nincluding the potential to assess features that may drastically impact\nclinical trial performance before a trial begins.\n\n\"Foundation models, combined with agentic workflows, will help unlock the full\npotential of precision medicine — enabling rapid hypothesis testing and\ndramatically compressing the time from discovery to clinical application,\"\nsaid Eric Lefkofsky, Founder and CEO of Tempus. \"The fact that our general\npurpose models are already outperforming highly tuned smaller models bodes\nwell for the ability of our novel biological multimodal foundation models to\nimprove clinical trial design and biomarker development, empowering physicians\nand life science companies in the pursuit of new diagnostics and novel\nmedicines.”\n\nAbout Tempus\n\nTempus is a technology company advancing precision medicine through the\npractical application of artificial intelligence in healthcare. With one of\nthe world’s largest libraries of multimodal data, and an operating system to\nmake that data accessible and useful, Tempus provides AI-enabled precision\nmedicine solutions to physicians to deliver personalized patient care and in\nparallel facilitates discovery, development and delivery of optimal\ntherapeutics. The goal is for each patient to benefit from the treatment of\nothers who came before by providing physicians with tools that learn as the\ncompany gathers more data. For more information, visit tempus.com\n(https://cts.businesswire.com/ct/CT?id=smartlink&url=http%3A%2F%2Ftempus.com%2F&esheet=54543854&newsitemid=20260529085034&lan=en-US&anchor=tempus.com&index=1&md5=ec6758d08b6840e294f8ce1ea7ae720a)\n.\n\nForward Looking Statements\n\nThis press release contains forward-looking statements within the meaning of\nSection 27A of the Securities Act of 1933, as amended (the “Securities\nAct”), and Section 21E of the Securities Exchange Act of 1934, as amended,\nabout Tempus and Tempus’ industry that involve substantial risks and\nuncertainties. All statements other than statements of historical facts\ncontained in this press release are forward-looking statements, including, but\nnot limited to, statements regarding expected outcomes and benefits of\nTempus’ suite of foundation models for insight generation into oncology. In\nsome cases, you can identify forward-looking statements because they contain\nwords such as “anticipate,” “believe,” “contemplate,”\n“continue,” “could,” “estimate,” “expect,” “going to,”\n“intend,” “may,” “plan,” “potential,” “predict,”\n“project,” “should,” “target,” “will,” or “would” or the\nnegative of these words or other similar terms or expressions. Tempus cautions\nyou that the foregoing may not include all of the forward-looking statements\nmade in this press release.\n\nYou should not rely on forward-looking statements as predictions of future\nevents. Tempus has based the forward-looking statements contained in this\npress release primarily on its current expectations and projections about\nfuture events and trends that it believes may affect Tempus’ business,\nfinancial condition, results of operations and prospects. These\nforward-looking statements are subject to risks and uncertainties related to:\nthe intended use of Tempus’ products and services; Tempus’ financial\nperformance; the ability to attract and retain customers and partners;\nmanaging Tempus’ growth and future expenses; competition and new market\nentrants; compliance with new laws, regulations and executive actions,\nincluding any evolving regulations in the artificial intelligence space; the\nability to maintain, protect and enhance Tempus’ intellectual property; the\nability to attract and retain qualified team members and key personnel; the\nability to repay or refinance outstanding debt, or to access additional\nfinancing; future acquisitions, divestitures or investments; the potential\nadverse impact of climate change, natural disasters, health epidemics,\nmacroeconomic conditions, and war or other armed conflict, as well as risks,\nuncertainties, and other factors described in the section titled “Risk\nFactors” in Tempus’ Annual Report on Form 10-K for the year ended December\n31, 2025, filed with the Securities and Exchange Commission (“SEC”) on\nFebruary 24, 2026, as well as in other filings Tempus may make with the SEC in\nthe future. In addition, any forward-looking statements contained in this\npress release are based on assumptions that Tempus believes to be reasonable\nas of this date. Tempus undertakes no obligation to update any forward-looking\nstatements to reflect events or circumstances after the date of this press\nrelease or to reflect new information or the occurrence of unanticipated\nevents, except as required by law.\n\n\n\nView source version on businesswire.com:\nhttps://www.businesswire.com/news/home/20260529085034/en/\n(https://www.businesswire.com/news/home/20260529085034/en/)\n\nHanah Heintzelman\n\nhanah.heintzelman@tempus.com (mailto:hanah.heintzelman@tempus.com)\n\n\nCopyright Business Wire 2026","article_body_html":"","raw_payload":{"data":{"id":"nBw1rVK10a","title":"Tempus Announces Initial Results from its Multimodal Foundation Model Efforts for Novel and Scalable Insight Generation in Oncology","author":"Business Wire","ticker":"TEM","created":"2026-05-29T13:15:00.172Z","tickers":["TEM"],"exchange":"NASDAQ","article_body":"Tempus Announces Initial Results from its Multimodal Foundation Model Efforts\nfor Novel and Scalable Insight Generation in Oncology\n\nTempus AI, Inc. (NASDAQ: TEM), a technology company leading the adoption of AI\nto advance precision medicine, today announced the latest results from its\nmission to build Multimodal Foundation Models at the 2026 American Society of\nClinical Oncology (ASCO) Annual Meeting. Tempus has been building novel\nbiological foundation models and agentic workflows by leveraging its more than\n500 petabytes of rich, molecularly grounded data—more than 45 million total\nde-identified patient journeys, 1.5 million with sequenced data, and over\n400,000 records in cancer with full genomic, transcriptomic, imaging, and\nclinical data. These efforts transform this data into unified patient\nrepresentations, unlocking actionable insights to speed precision medicine\nefforts in both the clinic and drug development.\n\nTempus’ latest multimodal, transformer-based model was trained on 2.5\nmillion longitudinal records encapsulating more than 250 million pages of\nclinical notes, 450,000 digitized medical images, and 500,000 genomic and\ntranscriptomic sequences. By aggregating modalities derived from\nbillion-parameter foundation models, it is designed to address thousands of\nprediction objectives anchored in overall survival (OS) and progression-free\nsurvival (PFS), without requiring additional data or model fine-tuning. This\narchitecture significantly reduces the time and data required to produce\nhundreds of clinically relevant insights for clinical trial design, patient\nrisk prediction, and novel multimodal diagnostics.\n\nIn a zero-shot setting without any further training, Tempus leveraged this\npatient trajectory model and a series of agentic workflows to unlock insights\nin a cohort of more than 1.2 million de-identified records with robust\nmultimodal data.\n\nAs a primary proof of concept, Tempus’ model was utilized to analyze\nEGFR-mutant NSCLC patients treated with osimertinib, the current frontline\nstandard of care third-generation EGFR-inhibitor. Tempus assessed whether the\nmodel could accurately stratify response to the standard of care treatment of\nosimertinib in patients with known clinically actionable biomarkers like EGFR.\nThis tests the model’s ability to learn a composite of features that predict\nresponse outside of known biological and clinical features to identify\npatients more likely to experience poor response to current therapies.\n\nWithout any pre-specified training, the model demonstrated:\n\n\n * Predictive Accuracy: A C-index of 0.802 for overall survival (p value <\n0.001).\n\n * Significant Survival Stratification: Hazard Ratio of 4.536 (95% CI: [3.289,\n6.255]) between high- and low-risk subgroups.\n\nThe model was evaluated against a range of mutation profiles (including more\nthan 30 EGFR-specific features including L858R, Exon 19 del); and the results\nstill produced prognostic value independent of molecular and clinical\nsubgroups, significantly stratifying the overall survival of TP53(+) patients\n(HR=5.96), as well as progression-free survival in patients without CNS\nmetastasis (HR=1.94). This outcome is just one example of how Tempus’ model\ncan produce novel clinical insights, as similar results have been observed for\npredicting overall survival in other patient populations.\n\nTempus’ multimodal patient trajectory models have also demonstrated\nsubstantial utility for drug development activities. For example, the models\nhave successfully predicted outcomes of patient cohorts that mirrored\npre-established clinical trials. In NSCLC, Tempus was able to assess cohorts\nmirroring those of practice-changing clinical trials (KEYNOTE-189, FLAURA-2,\nand DESTINY), and the multimodal patient trajectory model was able to\nzero-shot outperform standard approaches, such as Cox-PH modeling. These\noutcomes indicate substantial utility for biopharmaceutical developers,\nincluding the potential to assess features that may drastically impact\nclinical trial performance before a trial begins.\n\n\"Foundation models, combined with agentic workflows, will help unlock the full\npotential of precision medicine — enabling rapid hypothesis testing and\ndramatically compressing the time from discovery to clinical application,\"\nsaid Eric Lefkofsky, Founder and CEO of Tempus. \"The fact that our general\npurpose models are already outperforming highly tuned smaller models bodes\nwell for the ability of our novel biological multimodal foundation models to\nimprove clinical trial design and biomarker development, empowering physicians\nand life science companies in the pursuit of new diagnostics and novel\nmedicines.”\n\nAbout Tempus\n\nTempus is a technology company advancing precision medicine through the\npractical application of artificial intelligence in healthcare. With one of\nthe world’s largest libraries of multimodal data, and an operating system to\nmake that data accessible and useful, Tempus provides AI-enabled precision\nmedicine solutions to physicians to deliver personalized patient care and in\nparallel facilitates discovery, development and delivery of optimal\ntherapeutics. The goal is for each patient to benefit from the treatment of\nothers who came before by providing physicians with tools that learn as the\ncompany gathers more data. For more information, visit tempus.com\n(https://cts.businesswire.com/ct/CT?id=smartlink&url=http%3A%2F%2Ftempus.com%2F&esheet=54543854&newsitemid=20260529085034&lan=en-US&anchor=tempus.com&index=1&md5=ec6758d08b6840e294f8ce1ea7ae720a)\n.\n\nForward Looking Statements\n\nThis press release contains forward-looking statements within the meaning of\nSection 27A of the Securities Act of 1933, as amended (the “Securities\nAct”), and Section 21E of the Securities Exchange Act of 1934, as amended,\nabout Tempus and Tempus’ industry that involve substantial risks and\nuncertainties. All statements other than statements of historical facts\ncontained in this press release are forward-looking statements, including, but\nnot limited to, statements regarding expected outcomes and benefits of\nTempus’ suite of foundation models for insight generation into oncology. In\nsome cases, you can identify forward-looking statements because they contain\nwords such as “anticipate,” “believe,” “contemplate,”\n“continue,” “could,” “estimate,” “expect,” “going to,”\n“intend,” “may,” “plan,” “potential,” “predict,”\n“project,” “should,” “target,” “will,” or “would” or the\nnegative of these words or other similar terms or expressions. Tempus cautions\nyou that the foregoing may not include all of the forward-looking statements\nmade in this press release.\n\nYou should not rely on forward-looking statements as predictions of future\nevents. Tempus has based the forward-looking statements contained in this\npress release primarily on its current expectations and projections about\nfuture events and trends that it believes may affect Tempus’ business,\nfinancial condition, results of operations and prospects. These\nforward-looking statements are subject to risks and uncertainties related to:\nthe intended use of Tempus’ products and services; Tempus’ financial\nperformance; the ability to attract and retain customers and partners;\nmanaging Tempus’ growth and future expenses; competition and new market\nentrants; compliance with new laws, regulations and executive actions,\nincluding any evolving regulations in the artificial intelligence space; the\nability to maintain, protect and enhance Tempus’ intellectual property; the\nability to attract and retain qualified team members and key personnel; the\nability to repay or refinance outstanding debt, or to access additional\nfinancing; future acquisitions, divestitures or investments; the potential\nadverse impact of climate change, natural disasters, health epidemics,\nmacroeconomic conditions, and war or other armed conflict, as well as risks,\nuncertainties, and other factors described in the section titled “Risk\nFactors” in Tempus’ Annual Report on Form 10-K for the year ended December\n31, 2025, filed with the Securities and Exchange Commission (“SEC”) on\nFebruary 24, 2026, as well as in other filings Tempus may make with the SEC in\nthe future. In addition, any forward-looking statements contained in this\npress release are based on assumptions that Tempus believes to be reasonable\nas of this date. Tempus undertakes no obligation to update any forward-looking\nstatements to reflect events or circumstances after the date of this press\nrelease or to reflect new information or the occurrence of unanticipated\nevents, except as required by law.\n\n\n\nView source version on businesswire.com:\nhttps://www.businesswire.com/news/home/20260529085034/en/\n(https://www.businesswire.com/news/home/20260529085034/en/)\n\nHanah Heintzelman\n\nhanah.heintzelman@tempus.com (mailto:hanah.heintzelman@tempus.com)\n\n\nCopyright Business Wire 2026"},"type":"article","timestamp":"2026-05-29T13:15:00.298457962Z","server_sent_at_ms":1780060500298},"received_at":"2026-05-29T13:15:00.399Z","source_url":"https://www.businesswire.com/news/home/20260529085034/en/"},"analysis":{"id":"38837","press_release_id":"49406","analysis_json":{"industry":{"label":"Health Care Information Services","sector":"Health Care"},"redFlags":[],"eventType":"product_launch","narrative":"Tempus AI presented initial results from its multimodal foundation model at the ASCO Annual Meeting, leveraging over 500 petabytes of molecularly grounded data.\n\nThe model achieved a C-index of 0.802 for overall survival prediction in EGFR-mutant NSCLC patients treated with osimertinib, without any fine-tuning.\n\nIt demonstrated significant survival stratification with a hazard ratio of 4.536 between high- and low-risk subgroups, outperforming standard Cox-PH modeling in zero-shot settings.\n\nCEO Eric Lefkofsky stated that these general-purpose models outperform highly tuned smaller models, signaling progress in compressing the time from discovery to clinical application.","sentiment":"bullish","agentHooks":{"shouldPost":true,"suggestedAngle":"Tempus validates its AI moat with strong zero-shot predictive performance in oncology, reinforcing its precision medicine thesis."},"keyFigures":{"customDimensions":{"digitized_images":"450,000","test_cohort_size":"1.2 million","training_records":"2.5 million","clinical_notes_pages":"250 million","sequenced_data_records":"1.5 million","total_patient_journeys":"45 million","c_index_overall_survival":0.802,"hazard_ratio_tp53_positive":5.96,"hazard_ratio_high_vs_low_risk":4.536,"full_multimodal_cancer_records":"400,000","hazard_ratio_no_cns_metastasis":1.94,"genomic_transcriptomic_sequences":"500,000"}},"quotedText":"Foundation models, combined with agentic workflows, will help unlock the full potential of precision medicine — enabling rapid hypothesis testing and dramatically compressing the time from discovery to clinical application","namedEntities":{"people":[{"name":"Eric Lefkofsky","role":"Founder and CEO"},{"name":"Hanah Heintzelman","role":"IR Contact"}],"products":["Multimodal Foundation Model","osimertinib"],"companies":[{"name":"Tempus AI, Inc.","ticker":"TEM"},{"name":"American Society of Clinical Oncology","relationship":"conference host"}],"dollarAmounts":[]},"materialImpact":{"score":3,"reasoning":"Tempus announced initial technical results from its multimodal foundation model, demonstrating high predictive accuracy (C-index 0.802) and survival stratification in a key oncology cohort. This validates the company's core AI thesis and data moat, though it is a research milestone rather than a commercial revenue event."},"tickerRelevance":{"others":[],"primary":"TEM"},"globalImportance":45,"audienceRelevance":55,"eventTypeSecondary":["operations_update"],"importanceComponents":{"tickerTier":"mid-cap","eventGravity":"technical_milestone","sectorWeight":"high","householdBrandBoost":"low","retailFavoriteBoost":"moderate"}},"event_type":"product_launch","event_type_secondary":["operations_update"],"sentiment":"bullish","material_impact_score":3,"narrative":"Tempus AI presented initial results from its multimodal foundation model at the ASCO Annual Meeting, leveraging over 500 petabytes of molecularly grounded data.\n\nThe model achieved a C-index of 0.802 for overall survival prediction in EGFR-mutant NSCLC patients treated with osimertinib, without any fine-tuning.\n\nIt demonstrated significant survival stratification with a hazard ratio of 4.536 between high- and low-risk subgroups, outperforming standard Cox-PH modeling in zero-shot settings.\n\nCEO Eric Lefkofsky stated that these general-purpose models outperform highly tuned smaller models, signaling progress in compressing the time from discovery to clinical application.","key_figures":{"customDimensions":{"digitized_images":"450,000","test_cohort_size":"1.2 million","training_records":"2.5 million","clinical_notes_pages":"250 million","sequenced_data_records":"1.5 million","total_patient_journeys":"45 million","c_index_overall_survival":0.802,"hazard_ratio_tp53_positive":5.96,"hazard_ratio_high_vs_low_risk":4.536,"full_multimodal_cancer_records":"400,000","hazard_ratio_no_cns_metastasis":1.94,"genomic_transcriptomic_sequences":"500,000"}},"named_entities":{"people":[{"name":"Eric Lefkofsky","role":"Founder and CEO"},{"name":"Hanah Heintzelman","role":"IR Contact"}],"products":["Multimodal Foundation Model","osimertinib"],"companies":[{"name":"Tempus AI, Inc.","ticker":"TEM"},{"name":"American Society of Clinical Oncology","relationship":"conference host"}],"dollarAmounts":[]},"model_name":"qwen3_6_27b_awq","prompt_hash":"sha256:727b4b9429a443af","schema_hash":"sha256:05005c02d9cffac9","created_at":"2026-05-29T13:15:25.475Z","global_importance":45,"audience_relevance":55,"importance_components":{"tickerTier":"mid-cap","eventGravity":"technical_milestone","sectorWeight":"high","householdBrandBoost":"low","retailFavoriteBoost":"moderate"}},"durationMs":25019,"modelName":"george-droid-qwen-72b"}}