{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DDMY6GIIR7VE4D2EMSPCHIKBNG","short_pith_number":"pith:DDMY6GII","schema_version":"1.0","canonical_sha256":"18d98f19088fea4e0f44649e23a14169a600211a97a486a5071ab4184821424b","source":{"kind":"arxiv","id":"2506.04645","version":1},"attestation_state":"computed","paper":{"title":"Inference economics of language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Ege Erdil","submitted_at":"2025-06-05T05:28:09Z","abstract_excerpt":"We develop a theoretical model that addresses the economic trade-off between cost per token versus serial token generation speed when deploying LLMs for inference at scale. Our model takes into account arithmetic, memory bandwidth, network bandwidth and latency constraints; and optimizes over different parallelism setups and batch sizes to find the ones that optimize serial inference speed at a given cost per token. We use the model to compute Pareto frontiers of serial speed versus cost per token for popular language models."},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2506.04645","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-05T05:28:09Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"343372395adf22dd0a13c819876cea6d6c8b3d5e741a2472d6634103da8a0db1","abstract_canon_sha256":"4d71672d303eba2a680f9191c0a3fc761e1b27b2a95b793a51227909f7f46c4d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:18.434649Z","signature_b64":"eMG0jtdsyxCM1TSmgOMsK0yfDmCLEZ0AHnlaPrI5rdZ9w06ueRPm7Gn7sWYdNefgDeKRWMr6LsT7MkxhF78pBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"18d98f19088fea4e0f44649e23a14169a600211a97a486a5071ab4184821424b","last_reissued_at":"2026-07-05T11:16:18.434209Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:18.434209Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Inference economics of language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Ege Erdil","submitted_at":"2025-06-05T05:28:09Z","abstract_excerpt":"We develop a theoretical model that addresses the economic trade-off between cost per token versus serial token generation speed when deploying LLMs for inference at scale. Our model takes into account arithmetic, memory bandwidth, network bandwidth and latency constraints; and optimizes over different parallelism setups and batch sizes to find the ones that optimize serial inference speed at a given cost per token. We use the model to compute Pareto frontiers of serial speed versus cost per token for popular language models."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.04645","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2506.04645/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2506.04645","created_at":"2026-07-05T11:16:18.434267+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.04645v1","created_at":"2026-07-05T11:16:18.434267+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.04645","created_at":"2026-07-05T11:16:18.434267+00:00"},{"alias_kind":"pith_short_12","alias_value":"DDMY6GIIR7VE","created_at":"2026-07-05T11:16:18.434267+00:00"},{"alias_kind":"pith_short_16","alias_value":"DDMY6GIIR7VE4D2E","created_at":"2026-07-05T11:16:18.434267+00:00"},{"alias_kind":"pith_short_8","alias_value":"DDMY6GII","created_at":"2026-07-05T11:16:18.434267+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2606.11690","citing_title":"Beyond Per-Token Pricing: A Concurrency-Aware Methodology for LLM Infrastructure Cost Estimation","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02430","citing_title":"Not All Errors Are Equal: A Systematic Study of Error Propagation in Large Language Model Inference","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28680","citing_title":"A Techno-Economic Framework for Cost Modeling and Revenue Opportunities in Open and Programmable AI-RAN","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2509.22166","citing_title":"Motivating Next-Gen Accelerators with Flexible (N:M) Activation Sparsity via Benchmarking Lightweight Post-Training Sparsification Approaches","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28680","citing_title":"A Techno-Economic Framework for Cost Modeling and Revenue Opportunities in Open and Programmable AI-RAN","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02473","citing_title":"Analyzing Reverse Address Translation Overheads in Multi-GPU Scale-Up Pods","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01160","citing_title":"The Productivity-Reliability Paradox: Specification-Driven Governance for AI-Augmented Software Development","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08290","citing_title":"Tokalator: A Context Engineering Toolkit for Artificial Intelligence Coding Assistants","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG","json":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG.json","graph_json":"https://pith.science/api/pith-number/DDMY6GIIR7VE4D2EMSPCHIKBNG/graph.json","events_json":"https://pith.science/api/pith-number/DDMY6GIIR7VE4D2EMSPCHIKBNG/events.json","paper":"https://pith.science/paper/DDMY6GII"},"agent_actions":{"view_html":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG","download_json":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG.json","view_paper":"https://pith.science/paper/DDMY6GII","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.04645&json=true","fetch_graph":"https://pith.science/api/pith-number/DDMY6GIIR7VE4D2EMSPCHIKBNG/graph.json","fetch_events":"https://pith.science/api/pith-number/DDMY6GIIR7VE4D2EMSPCHIKBNG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG/action/storage_attestation","attest_author":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG/action/author_attestation","sign_citation":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG/action/citation_signature","submit_replication":"https://pith.science/pith/DDMY6GIIR7VE4D2EMSPCHIKBNG/action/replication_record"}},"created_at":"2026-07-05T11:16:18.434267+00:00","updated_at":"2026-07-05T11:16:18.434267+00:00"}