Back to AI News
arxiv

Sable: A Sumcheck Proof System for Bit-Exact LLM Inference

Sable is an argument system and single-GPU prover that proves a committed, integer-quantized transformer produced a given token sequence from a given prompt bit for bit. It fixes an integer-only function with deterministic rounding, normalization, softmax and sampling, then proves it using batched zerochecks, LogUp-GKR lookups, multilinear commitments, Reed-Solomon codes, BLAKE3 Merkle trees and a mixed-length BaseFold opening. On one NVIDIA RTX PRO 6000 Blackwell GPU, the prover covers all 64 layers of Qwen2.5-32B in 92.5 seconds for 4,096 tokens and 398 seconds for 16,384 tokens, and also covers Qwen3-32B and Qwen3-30B-A3B. The proof costs 41 to 50 times serving time, or 6.4 cents per 4,096-token request at public GPU prices; a verifier checks a proof on one CPU core in 19 to 56 seconds, and the system is not zero-knowledge.

AI Brief

It gives concrete performance and cost data for cryptographic verification of open-weight LLM inference, a useful development for trust and auditability research.

arXiv:2611.14083v1 Announce Type: cross Abstract: We present Sable, an argument system and single-GPU prover for the statement that a committed, integer-quantized transformer produced a given token sequence from a given prompt, bit for bit. Sable fixes an integer function F_O with no floating-point step, in which every rounding, normalization, softmax and sampling decision is determined by its inputs, and proves it with a constraint system whose relations provably never wrap modulo the 31-bit KoalaBear prime. The proof batches zerochecks and LogUp-GKR lookups into one sumcheck per domain, and commits to every witness tensor and weight in one multilinear commitment built from Reed-Solomon codes, BLAKE3 Merkle trees and a mixed-length BaseFold opening. The prover streams the model layer by layer on one GPU. On one NVIDIA RTX PRO 6000 Blackwell GPU, Sable proves all 64 layers of Qwen2.5-32B in 92.5 s for 4,096 tokens and 398 s for 16,384 tokens, and covers Qwen3-32B and the mixture-of-experts Qwen3-30B-A3B (4,096 tokens in 121.8 and 84.7 s). A proof costs 41 to 50 times the serving time, or 6.4 cents per 4,096-token request at public GPU prices, and 40 to 64 times less time per parameter and token than zkLLM. On WikiText-2, F_O's greedy token matches the BF16 model's at 86.3% of positions; on Qwen3, F_O is within 1.6 points of BF16 on MMLU and 1.7 points on GSM8K. A verifier with a compiled core checks a proof on one CPU core in 19 to 56 s. In the random-oracle model, the interactive proof has round-by-round soundness error 2^-103. The proof is not zero-knowledge, so the system targets open-weight models. Code: https://github.com/moonmath-ai/Sable
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!