QuanBench+ - PennyLane: leaderboard

Quantum code generation in PennyLane on QuanBench+ (2026): 42 tasks on quantum algorithms, gate decomposition and state preparation, aligned across Qiskit, Cirq and PennyLane and checked by executable functional tests (probabilistic outputs accepted by KL divergence). Pass@1 with greedy decoding, in percent; higher is better.

Metric: Pass@1 (%): greedy decoding at temperature 0, one PennyLane program per task, on the 42 framework-aligned QuanBench+ tasks (quantum algorithms, gate decomposition, state preparation), graded by executable functional tests with KL-divergence acceptance for probabilistic outputs; higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.142.9#248
2Gemini 3 Pro40.5#146
3DeepSeek R133.3#484
4MiniMax-M2.131#582
5QuanBench+ DeepSeek-Chat (checkpoint unspecified)31
6Claude 3.7 Sonnet26.2#455
7Kimi K2 (Thinking)26.2
8Gemini 2.5 Flash23.8#466
9GPT-4.123.8#506
10GLM-4.723.8#383
11Llama 4 Maverick19#1113
12Qwen 2.5 7B Instruct11.9#2732

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=quanbench-plus-pennylane · How It Works · Data refreshed daily, snapshot 2026-11-22.