Local Agent Bench - A100 SXM4 80 GB - Three Trials - Injection Resistance: leaderboard
Local Agent Bench v0.1.4, Injection Resistance native percent on the public 100-task suite, three trials for each of five exact Ollama GGUF artifacts on A100-SXM4-80GB. Scoring v3; useful-and-safe injection utility, Hermes Agent and Ollama0.40.6, actual loaded context32768. Separate hardware population, never pooled with RTX3090/M2/RTX3070; speed/cost are not the score.
Metric: Injection Resistance (%). Source: localagentbench.com. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Ollama gemma4:31b-it-q8_0 (Local Agent Bench Run: quant=Q8_0;gguf=604816c521aa780a6b61841f03d4598002feb2d8c73b3aefaf4861d17461dfec;engine=ollama0.40.6;reasoning=default;trials=3) | 100 | |
| 2 | Ollama qwen2.5:72b (Local Agent Bench Run: quant=Q4_K_M;gguf=6e7fdda508e91cb0f63de5c15ff79ac63a1584ccafd751c07ca12b7f442101b8;engine=ollama0.40.6;reasoning=default;trials=3) | 100 | |
| 3 | Ollama llama3.3:70b (Local Agent Bench Run: quant=Q4_K_M;gguf=4824460d29f2058aaf6e1118a63a7a197a09bed509f0e7d4e2efb1ee273b447d;engine=ollama0.40.6;reasoning=default;trials=3) | 100 | |
| 4 | Ollama deepseek-r1:70b (Local Agent Bench Run: quant=Q4_K_M;gguf=4cd576d9aa16961244012223abf01445567b061f1814b57dfef699e4cf8df339;engine=ollama0.40.6;reasoning=default;trials=3) | 100 | |
| 5 | Ollama qwen3:32b-q8_0 (Local Agent Bench Run: quant=Q8_0;gguf=de447d788da3df6b4ea340408b13fc2c3a2043a2dfc19178b12d501a4bd96484;engine=ollama0.40.6;reasoning=default;trials=3) | 73.33 |
Interactive version: theaggregate.ai/benchmark?slug=local-agent-bench-a100-sxm4-80-gb-three-trials-injection-resistance · How It Works · Data refreshed daily, snapshot 2026-11-12.