# SysBench: system message following (SysBench)

> Tests how well a model follows the rules in a system message across single and multi-turn conversations.

Canonical page: https://benchlm.ai/benchmarks/sysbench

- Category: [Instruction Following](/instruction-following)
- Last updated: October 29, 2026

## About SysBench

- Year: 2024
- Tasks: System-message constraints over conversations
- Format: Instruction success rate
- Difficulty: Keeping long prompt rules
- Paper: [SysBench: Can Large Language Models Follow System Messages?](https://github.com/PKU-Baichuan-MLSystemLab/SysBench)

NVIDIA reports it for Nemotron 3.5 Bolt E4B (October 23, 2026), run in its NeMo Gym harness, as the instruction success rate in reasoning mode.

SysBench is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (2 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Nemotron 3.5 Bolt E4B BF16](/models/nemotron-3-5-bolt-e4b-bf16) | NVIDIA | 67.88% |
| 2 | [Nemotron 3.5 Bolt E4B](/models/nemotron-3-5-bolt-e4b) | NVIDIA | 67.22% |

## FAQ

### What does SysBench measure?

Tests how well a model follows the rules in a system message across single and multi-turn conversations.

### Which model scores highest on SysBench?

Nemotron 3.5 Bolt E4B BF16 by NVIDIA currently leads with a score of 67.88% on SysBench.

### How many models are evaluated on SysBench?

2 AI models have published results on SysBench in the BenchLM catalog.

### Does SysBench affect BenchLM's overall score?

Not directly. SysBench is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on SysBench

- [Nemotron 3.5 Bolt E4B BF16 vs Nemotron 3.5 Bolt E4B](/compare/nemotron-3-5-bolt-e4b-vs-nemotron-3-5-bolt-e4b-bf16)