ASI Prize

About

The only benchmark where the problems are unsolved and the verification is zero-trust. The Lean 4 compiler is the sole judge of proof correctness.

ASI Prize mission: benchmarking AI against unsolved mathematics with zero-trust Lean 4 verification
Our Mission

Track the rise of superintelligent reasoning.

300 conjectures no human has proven, formalized in Lean 4. Either the proof type-checks or it doesn't. No partial credit, no ambiguity, no human judge. A machine decides.

Why This Exists

Existing benchmarks use solved textbook problems where answers are known. We use genuinely unsolved conjectures - the same problems professional mathematicians haven't cracked.

If an AI proves one of these theorems, the proof is independently verifiable by the Lean 4 type-checker. No human review required. That's zero-trust verification.

The Benchmark

300 pure problems filtered from 1,482 open conjectures in Google DeepMind's formal-conjectures repo. Covers number theory, combinatorics, analysis, algebra, and more.

300
Benchmark Problems
Lean 4.27.0
Theorem Prover
Mathlib v4.27.0
7,735 Modules
19
Banned Tokens

Zero-Trust Verification

We recently transitioned from a naive V1 pipeline to a robust V2 architecture centered around Lean's official Comparator for true zero-trust verification, preventing adversarial redefinition attacks.

  • Banned token scan - Pre-scan for 19 tokens (sorry, admit, native_decide, etc.)
  • Comparator execution - Runs a sandboxed kernel replay of the compiled proof.
  • Exact signature match - Checks that the theorem signature perfectly matches the ground truth.
  • Axiom audit - Validates that no unauthorized axioms (like sorryAx) are injected into the kernel environment.
  • Fail-safe - Enforced via systemd-run scopes preventing resource exhaustion attacks.

Problem Sources

  • Erdos Problems - Classic combinatorics and number theory
  • Green's Open Problems - Additive combinatorics
  • Arxiv Conjectures - Recent research papers formalized
  • OEIS Sequences - Integer sequence conjectures
  • Millennium Problems - Clay Math Institute prizes

Scoring

Verified: The proof compiles with zero warnings, zero banned tokens, and clean axioms. Binary pass/fail. This is the only metric that matters.

Progress: Goals solved divided by total goals. Each problem is one theorem with one top-level goal. If any goals remain unsolved, progress is 0%.

Primary ranking is by verified count. No partial credit.

Current Results

4 AI models evaluated with up to 10 iterations of compiler feedback per problem. 1 verified proof so far: Green's Problem 24 (upper_trivial, a trivial upper bound, formally verified), proved by Gemini 3 Flash in 178 seconds. $2.78M in prize money remains unclaimed across 12 problems.

Mathematics prize conjectures from Clay Math, Erdos, and other foundations
Prize Challenges

Solve an Open Problem

From $10 Erdos problems to $1M Millennium Prizes. Prizes administered by their respective organizations.

View prize problems →
AI model leaderboard for formal mathematics proof verification
Model Evaluation

Leaderboard

See which AI models can prove unsolved conjectures. Per-model analysis, error breakdowns, and progress charts.

View leaderboard →