{"paper":{"title":"Mistral 7B","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A 7-billion-parameter model outperforms Llama 2 13B on every benchmark and Llama 1 34B on reasoning, mathematics, and code generation.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, L\\'elio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timoth\\'ee Lacroix, William El Sayed","submitted_at":"2023-10-10T17:54:58Z","abstract_excerpt":"We introduce Mistral 7B v0.1, a 7-billion-parameter language model engineered for superior performance and efficiency. Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation. Our model leverages grouped-query attention (GQA) for faster inference, coupled with sliding window attention (SWA) to effectively handle sequences of arbitrary length with a reduced inference cost. We also provide a model fine-tuned to follow instructions, Mistral 7B -- Instruct, that surpasses the Llama 2 13B -- Chat model both on human and auto"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The chosen evaluation benchmarks and test sets are representative of downstream usefulness and contain no data leakage from the training corpus.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Mistral 7B is a 7B-parameter LLM that outperforms Llama 2 13B across benchmarks via grouped-query attention and sliding-window attention while remaining efficient.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A 7-billion-parameter model outperforms Llama 2 13B on every benchmark and Llama 1 34B on reasoning, mathematics, and code generation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a254c6d86f39e2ac32174c4b17c307c7ca802e3fdc325d332906e09648ca7f04"},"source":{"id":"2310.06825","kind":"arxiv","version":1},"verdict":{"id":"60e91080-3564-4839-bf6a-55d1d74e61e6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-24T06:09:42.492805Z","strongest_claim":"Mistral 7B outperforms Llama 2 13B across all evaluated benchmarks, and Llama 1 34B in reasoning, mathematics, and code generation.","one_line_summary":"Mistral 7B is a 7B-parameter LLM that outperforms Llama 2 13B across benchmarks via grouped-query attention and sliding-window attention while remaining efficient.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The chosen evaluation benchmarks and test sets are representative of downstream usefulness and contain no data leakage from the training corpus.","pith_extraction_headline":"A 7-billion-parameter model outperforms Llama 2 13B on every benchmark and Llama 1 34B on reasoning, mathematics, and code generation."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2310.06825/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":29,"sample":[{"doi":"","year":2023,"title":"GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints","work_id":"b73ad5b2-e553-4c71-b0c9-67e67ba7b158","ref_index":1,"cited_arxiv_id":"2305.13245","is_internal_anchor":true},{"doi":"","year":2021,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","ref_index":2,"cited_arxiv_id":"2108.07732","is_internal_anchor":true},{"doi":"","year":2004,"title":"Longformer: The Long-Document Transformer","work_id":"abea7a44-6668-4de7-aab6-f53a6e5aa088","ref_index":3,"cited_arxiv_id":"2004.05150","is_internal_anchor":true},{"doi":"","year":2020,"title":"Piqa: Reasoning about phys- ical commonsense in natural language","work_id":"7cb4a126-31c8-4ff4-bf34-12bd8bdc74fe","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","ref_index":5,"cited_arxiv_id":"2107.03374","is_internal_anchor":true}],"resolved_work":29,"snapshot_sha256":"af5cbfc51579651f8062a6608da657e25ecbc7e32801994732e526a7da1cf952","internal_anchors":21},"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"}