{"id":"https://openalex.org/W7203682669","doi":"https://doi.org/10.48550/arxiv.2608.16157","title":"FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution","display_name":"FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution","publication_year":2026,"publication_date":"2026-08-17","ids":{"openalex":"https://openalex.org/W7203682669","doi":"https://doi.org/10.48550/arxiv.2608.16157"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.16157","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.16157","pdf_url":"https://arxiv.org/pdf/2608.16157","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.16157","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5147444645","display_name":"Shuo Yang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Shuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128767762","display_name":"Xiaoze Fan","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fan, Xiaoze","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136800963","display_name":"Melissa Pan","orcid":"https://orcid.org/0009-0001-7826-2049","observed_orcids":["https://orcid.org/0009-0001-7826-2049"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pan, Melissa","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147423797","display_name":"Haocheng Xi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xi, Haocheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5159818512","display_name":"Zhe Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100606010","display_name":"Shanlin Sun","orcid":"https://orcid.org/0000-0002-4519-7287","observed_orcids":["https://orcid.org/0000-0002-4519-7287"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Shanlin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047285420","display_name":"Kurt Keutzer","orcid":"https://orcid.org/0000-0003-3868-8501","observed_orcids":["https://orcid.org/0000-0003-3868-8501"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Keutzer, Kurt","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070926896","display_name":"Song Han","orcid":"https://orcid.org/0000-0002-4186-7618","observed_orcids":["https://orcid.org/0000-0002-4186-7618"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Song","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005554337","display_name":"Matei Zaharia","orcid":"https://orcid.org/0000-0002-7547-7204","observed_orcids":["https://orcid.org/0000-0002-7547-7204"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zaharia, Matei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147415565","display_name":"Chenfeng Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Chenfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5041920173","display_name":"Ion Gabriel Stoica","orcid":"https://orcid.org/0000-0002-5373-0088","observed_orcids":["https://orcid.org/0000-0002-5373-0088"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Stoica, Ion","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":true,"cited_by_count":10,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4447999894618988,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4447999894618988,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.17870000004768372,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.16940000653266907,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/mixture-of-experts","display_name":"mixture of experts","score":0.6639999747276306},{"id":"https://openalex.org/keywords/open-weight-models","display_name":"open-weight models","score":0.7329999804496765},{"id":"https://openalex.org/keywords/local-large-language-models","display_name":"local large language models","score":0.5120000243186951},{"id":"https://openalex.org/keywords/tool-using-agents","display_name":"tool-using agents","score":0.4860000014305115}],"concepts":[{"id":"https://openalex.org/C2780008327","wikidata":"https://www.wikidata.org/wiki/Q3962","display_name":"Laptop","level":2,"score":0.883400022983551},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8151000142097473},{"id":"https://openalex.org/C67953723","wikidata":"https://www.wikidata.org/wiki/Q192525","display_name":"Workstation","level":2,"score":0.6953999996185303},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.6118999719619751},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.5544999837875366},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4925000071525574},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.39259999990463257},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.36419999599456787},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.36000001430511475},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.3582000136375427},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.34769999980926514},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.3395000100135803},{"id":"https://openalex.org/C25344961","wikidata":"https://www.wikidata.org/wiki/Q192726","display_name":"Virtual machine","level":2,"score":0.32109999656677246},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.3165999948978424},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.313400000333786},{"id":"https://openalex.org/C2776252893","wikidata":"https://www.wikidata.org/wiki/Q1364836","display_name":"FLEX","level":2,"score":0.29409998655319214},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.2759000062942505},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.27250000834465027},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C184596265","wikidata":"https://www.wikidata.org/wiki/Q2651576","display_name":"Model of computation","level":3,"score":0.25780001282691956},{"id":"https://openalex.org/C32833848","wikidata":"https://www.wikidata.org/wiki/Q4115054","display_name":"Extensibility","level":2,"score":0.25699999928474426},{"id":"https://openalex.org/C70388272","wikidata":"https://www.wikidata.org/wiki/Q5968558","display_name":"IBM","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.16157","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.16157","pdf_url":"https://arxiv.org/pdf/2608.16157","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.16157","doi":"https://doi.org/10.48550/arxiv.2608.16157","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.16157","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.16157","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.16157","pdf_url":"https://arxiv.org/pdf/2608.16157","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7203682669.pdf","grobid_xml":"https://content.openalex.org/works/W7203682669.grobid-xml"},"referenced_works_count":27,"referenced_works":["https://openalex.org/W4321636575","https://openalex.org/W4324297016","https://openalex.org/W4387321091","https://openalex.org/W4388092923","https://openalex.org/W4390438324","https://openalex.org/W4391272939","https://openalex.org/W4391800686","https://openalex.org/W4404308452","https://openalex.org/W4404341652","https://openalex.org/W4404401018","https://openalex.org/W4405254248","https://openalex.org/W4406032994","https://openalex.org/W4408017080","https://openalex.org/W4414198736","https://openalex.org/W4415333510","https://openalex.org/W4415797413","https://openalex.org/W4416447653","https://openalex.org/W7075994024","https://openalex.org/W7103188877","https://openalex.org/W7103754783","https://openalex.org/W7105198942","https://openalex.org/W7150801976","https://openalex.org/W7155538430","https://openalex.org/W7165640641","https://openalex.org/W7166707533","https://openalex.org/W7171552453","https://openalex.org/W7172396375"],"related_works":["https://openalex.org/W4414490806","https://openalex.org/W7165900867","https://openalex.org/W7169841475","https://openalex.org/W4407759190","https://openalex.org/W7223129074","https://openalex.org/W4410986266","https://openalex.org/W7135023852","https://openalex.org/W7214073840","https://openalex.org/W7165649855","https://openalex.org/W7154972415"],"abstract_inverted_index":{"Frontier":[0],"open-weight":[1],"models":[2,113],"are":[3],"increasingly":[4],"available,":[5],"but":[6,32],"serving":[7,20,43],"them":[8],"still":[9],"largely":[10],"assumes":[11],"datacenter":[12],"infrastructure.":[13],"We":[14,187],"present":[15],"FreeToken,":[16],"an":[17,124],"edge-native":[18],"MoE":[19,112],"system":[21,190],"that":[22],"treats":[23],"a":[24,29,34,91,129,144,148,151,155,163,181],"personal":[25],"machine":[26,84],"not":[27],"as":[28,33],"small":[30],"GPU,":[31],"unified,":[35],"elastic":[36],"inference":[37],"platform.":[38],"FreeToken":[39,95,107,167],"co-designs":[40],"the":[41,103,159,176,189],"full":[42],"stack,":[44],"including":[45],"model":[46,100,146,153],"layout":[47],"and":[48,57,74,99,114,117,158],"loading,":[49],"expert":[50],"residency,":[51],"CPU--GPU":[52],"execution,":[53],"agentic":[54],"state":[55,101],"reuse,":[56],"runtime":[58],"memory":[59],"management,":[60],"around":[61],"two":[62],"realities":[63],"of":[64],"local":[65,173],"AI:":[66],"agent":[67],"workloads":[68],"continuously":[69,96],"change":[70],"their":[71],"execution":[72],"pattern,":[73],"edge":[75],"hardware":[76,121],"exposes":[77],"heterogeneous":[78],"resources":[79,104],"whose":[80],"balance":[81],"differs":[82],"from":[83,123,143],"to":[85,90,128,150],"machine.":[86],"Rather":[87],"than":[88,110],"committing":[89],"fixed":[92],"offloading":[93],"strategy,":[94],"maps":[97],"computation":[98],"onto":[102],"actually":[105],"available.":[106],"supports":[108],"more":[109],"20":[111],"real":[115],"coding":[116],"tool-using":[118],"agents":[119],"across":[120],"ranging":[122],"8GB":[125],"laptop":[126,149],"GPU":[127],"single":[130,164],"workstation":[131,165],"GPU.":[132,166],"More":[133],"importantly,":[134],"it":[135],"changes":[136],"what":[137],"these":[138],"machines":[139,177],"can":[140],"practically":[141],"serve,":[142],"35B":[145],"on":[147,154,162],"284B":[152],"gaming":[156],"desktop":[157],"753B":[160],"GLM-5.2":[161],"turns":[168],"open":[169],"weights":[170],"into":[171],"deployable":[172],"software,":[174],"making":[175],"users":[178],"already":[179],"own":[180],"practical":[182],"platform":[183],"for":[184],"frontier-scale":[185],"intelligence.":[186],"release":[188],"at":[191],"flashml.ai.":[192]},"counts_by_year":[{"year":2026,"cited_by_count":10}],"updated_date":"2026-11-09T13:22:03.289127","created_date":"2026-08-19T00:00:00"}
