{"id":"https://openalex.org/W7160268707","doi":"https://doi.org/10.48550/arxiv.2605.01158","title":"The Hidden Cost of Thinking: Energy Use and Environmental Impact of LMs Beyond Pretraining","display_name":"The Hidden Cost of Thinking: Energy Use and Environmental Impact of LMs Beyond Pretraining","publication_year":2026,"publication_date":"2026-05-01","ids":{"openalex":"https://openalex.org/W7160268707","doi":"https://doi.org/10.48550/arxiv.2605.01158"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.01158","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01158","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.01158","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135401248","display_name":"Jacob Morrison","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Morrison, Jacob","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135367978","display_name":"Noah A. Smith","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Smith, Noah A.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5051203366","display_name":"Emma Strubell","orcid":"https://orcid.org/0000-0003-2798-0726"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Strubell, Emma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.1152999997138977,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.1152999997138977,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.07800000160932541,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11948","display_name":"Machine Learning in Materials Science","score":0.07530000060796738,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5927000045776367},{"id":"https://openalex.org/keywords/environmental-impact-assessment","display_name":"Environmental impact assessment","score":0.5389000177383423},{"id":"https://openalex.org/keywords/pipeline-transport","display_name":"Pipeline transport","score":0.5152999758720398},{"id":"https://openalex.org/keywords/energy-consumption","display_name":"Energy consumption","score":0.4916999936103821},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.47380000352859497},{"id":"https://openalex.org/keywords/consumption","display_name":"Consumption (sociology)","score":0.4625999927520752},{"id":"https://openalex.org/keywords/energy","display_name":"Energy (signal processing)","score":0.3416000008583069},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.30329999327659607}],"concepts":[{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5927000045776367},{"id":"https://openalex.org/C164749845","wikidata":"https://www.wikidata.org/wiki/Q320389","display_name":"Environmental impact assessment","level":2,"score":0.5389000177383423},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.5152999758720398},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.4916999936103821},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.47380000352859497},{"id":"https://openalex.org/C30772137","wikidata":"https://www.wikidata.org/wiki/Q5164762","display_name":"Consumption (sociology)","level":2,"score":0.4625999927520752},{"id":"https://openalex.org/C134560507","wikidata":"https://www.wikidata.org/wiki/Q753291","display_name":"Environmental economics","level":1,"score":0.45419999957084656},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4528000056743622},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.36399999260902405},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3416000008583069},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C2776356880","wikidata":"https://www.wikidata.org/wiki/Q5381206","display_name":"Environmental data","level":2,"score":0.30140000581741333},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.2874000072479248},{"id":"https://openalex.org/C2984118289","wikidata":"https://www.wikidata.org/wiki/Q29954","display_name":"Power consumption","level":3,"score":0.2849999964237213},{"id":"https://openalex.org/C107826830","wikidata":"https://www.wikidata.org/wiki/Q929380","display_name":"Environmental resource management","level":1,"score":0.27730000019073486},{"id":"https://openalex.org/C539469273","wikidata":"https://www.wikidata.org/wiki/Q1749732","display_name":"Environmental monitoring","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.26899999380111694},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.2653999924659729},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.26489999890327454},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.2639000117778778},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C39432304","wikidata":"https://www.wikidata.org/wiki/Q188847","display_name":"Environmental science","level":0,"score":0.2596000134944916},{"id":"https://openalex.org/C2778136018","wikidata":"https://www.wikidata.org/wiki/Q10350689","display_name":"Predictive power","level":2,"score":0.25189998745918274}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.01158","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01158","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.01158","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01158","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Modern":[0],"language":[1],"model":[2,39,130,170],"development":[3,40,131],"extends":[4],"far":[5],"beyond":[6],"pretraining,":[7],"yet":[8],"environmental":[9,34,187,199],"reporting":[10,188],"remains":[11],"narrowly":[12],"focused":[13],"on":[14],"the":[15,28,33,116,192],"cost":[16],"of":[17,32,36,57,88,108,136,146],"training":[18],"a":[19,37,55,111],"single":[20],"final":[21],"model.":[22],"In":[23,125],"this":[24],"work,":[25],"we":[26,127],"provide":[27],"first":[29],"detailed":[30],"breakdown":[31],"impact":[35],"full":[38],"pipeline,":[41],"from":[42],"pretraining":[43],"through":[44],"supervised":[45],"fine-tuning,":[46],"preference":[47],"optimization,":[48],"and":[49,60,68,103,142,181,190],"reinforcement":[50,93],"learning,":[51],"for":[52,106,119,185],"Olmo":[53],"3,":[54],"family":[56],"7":[58],"billion":[59,62],"32":[61],"parameter":[63],"models":[64,75],"in":[65,86,122,186],"both":[66],"instruction-following":[67],"reasoning":[69,74],"variants.":[70],"We":[71],"find":[72],"that":[73],"are":[76,165,172],"17x":[77],"more":[78,179],"expensive":[79],"to":[80,196],"post-train":[81],"than":[82,158],"their":[83],"instruction-tuned":[84],"counterparts":[85],"terms":[87],"datacenter":[89,137],"energy,":[90,138],"driven":[91,151],"by":[92,153,169,191],"learning":[94],"rollout":[95],"generation.":[96],"Development":[97],"costs":[98],"(including":[99],"experimentation,":[100],"failed":[101],"runs,":[102],"ablations)":[104],"account":[105],"82.2%":[107],"total":[109],"compute,":[110],"roughly":[112],"65%":[113],"increase":[114],"over":[115],"~50%":[117],"reported":[118],"pretraining-focused":[120],"pipelines":[121,177],"prior":[123],"work.":[124],"total,":[126],"estimate":[128],"our":[129],"process":[132],"consumed":[133,143],"~12.3":[134],"GWh":[135],"emitted":[139],"4,251":[140],"tCO2eq,":[141],"15,887":[144],"kL":[145],"water,":[147],"with":[148],"water":[149],"consumption":[150],"entirely":[152,167],"power":[154],"generation":[155],"infrastructure":[156],"rather":[157],"data":[159],"center":[160],"cooling.":[161],"These":[162],"costs,":[163],"which":[164],"almost":[166],"unreported":[168],"developers,":[171],"growing":[173],"rapidly":[174],"as":[175],"post-training":[176],"become":[178],"complex,":[180],"must":[182],"be":[183],"accounted":[184],"standards":[189],"research":[193],"community":[194],"working":[195],"reduce":[197],"AI's":[198],"impact.":[200]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-06T00:00:00"}