{"slug": "axiomprover-at-imo-2026-perfect-score", "title": "AxiomProver at IMO 2026 (perfect score)", "summary": "AxiomProver, an autonomous multi-agent theorem prover for Lean 4 developed by Axiom Math, achieved a perfect score of 42/42 at the 2026 International Mathematical Olympiad in Shanghai, solving all six problems. The verified formal proofs are publicly available on GitHub, marking a milestone in AI-driven mathematical reasoning.", "body_md": "IMO 2026, the world's most prestigious pre-college mathematics competition, was\nheld in Shanghai on July 15–16, 2026. AxiomProver solved all six problems, achieving a\nperfect score of 42/42. AxiomProver is an autonomous multi-agent ensemble theorem\nprover for Lean 4, developed by [Axiom Math](https://axiommath.ai/).\n\nThis repository contains the formal Lean 4 statements and solutions for all six problems of the competition.\n\nThe official source of the problems is [https://www.imo-official.org/problems/2026/](https://www.imo-official.org/problems/2026/).\n\nEach problem lives under `IMO2026/<code>/`\n\n:\n\n`problem.lean`\n\n— the formal statement, with the bodies left as`sorry`\n\n, autonomously generated by AxiomProver.`solution.lean`\n\n— the verified formal proof, autonomously generated by AxiomProver.\n\n**2026 Q1**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q1/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q1/solution.lean)(521 lines, 24 minutes).**2026 Q2**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q2/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q2/solution.lean)(1224 lines, 360 minutes).**2026 Q3**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q3/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q3/solution.lean)(4229 lines, 869 minutes).**2026 Q4**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q4/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q4/solution.lean)(520 lines, 39 minutes).**2026 Q5**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q5/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q5/solution.lean)(457 lines, 65 minutes).**2026 Q6**:[[statement]](/AxiomMath/IMO2026/blob/main/IMO2026/Q6/problem.lean)[[solution]](/AxiomMath/IMO2026/blob/main/IMO2026/Q6/solution.lean)(771 lines, 139 minutes).\n\nBuilt against Mathlib `v4.31.0`\n\n(see `lean-toolchain`\n\n).\n\n```\nlake exe cache get   # fetch the prebuilt Mathlib cache\nlake build           # build all problem/solution libraries\n```\n\nOne can verify that each `problem.lean`\n\nand `solution.lean`\n\nare compatible\nusing `verify.py`\n\n, which calls [Axle's verify_proof](https://axle.axiommath.ai/):\n\n```\npython3 verify.py\nQ1: okay=True (passed)\nQ2: okay=True (passed)\nQ3: okay=True (passed)\nQ4: okay=True (passed)\nQ5: okay=True (passed)\nQ6: okay=True (passed)\n```\n\nThis is expected to complete very quickly, as the results are cached by Axle.\nTo bypass this, pass `--no-cache`\n\nto the call, which will force Axle to recompute everything,\nat the cost of a slower time:\n\n```\npython3 verify.py --no-cache\nQ1: okay=True (passed)\nQ2: okay=True (passed)\nQ3: okay=True (passed)\nQ4: okay=True (passed)\nQ5: okay=True (passed)\nQ6: okay=True (passed)\n```\n\n", "url": "https://wpnews.pro/news/axiomprover-at-imo-2026-perfect-score", "canonical_source": "https://github.com/AxiomMath/IMO2026", "published_at": "2026-07-17 09:53:29+00:00", "updated_at": "2026-07-17 10:21:37.927071+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-research", "ai-products", "ai-agents"], "entities": ["AxiomProver", "Axiom Math", "Lean 4", "IMO 2026", "Mathlib", "Axle"], "alternates": {"html": "https://wpnews.pro/news/axiomprover-at-imo-2026-perfect-score", "markdown": "https://wpnews.pro/news/axiomprover-at-imo-2026-perfect-score.md", "text": "https://wpnews.pro/news/axiomprover-at-imo-2026-perfect-score.txt", "jsonld": "https://wpnews.pro/news/axiomprover-at-imo-2026-perfect-score.jsonld"}}