Everyone is building agents, but it seems like everyone is missing the performance side of agent deployments.
I built AutoAgents, an AI agent framework in Rust. Not because Rust is trendy — because I was tired of paying for framework overhead that added zero value, and not having enough modularity to modify agent executors or memory.
I benchmarked 7 AI frameworks. Same machine. Same LLM. Same prompts. Same concurrency. No tricks.
I wanted to give users the full picture on what frameworks to choose given the situation.
Comparing Rust to Python isn't an apples-to-apples comparison, but the goal was to give a clear picture of how the framework affects resource usage so users can make informed decisions. The tools and async layer are built on Pandas and Asyncio to ensure better Python performance.
We're not saying Python agent frameworks are bad. They got this ecosystem started and the DX is genuinely great. We're saying that at production scale, the runtime cost of convenience becomes a real line item — and teams deserve to see those numbers before they commit.
A small peek into the memory usage:
Rust frameworks: ~1 GB at 10 sessions
Python frameworks: 5–9 GB at 10 sessions
The benchmark is released on GitHub so you can explore every metric yourself — latency distributions, P95/P99 tails, CPU, memory.
AutoAgents Repo: https://github.com/liquidos-ai/AutoAgents Benchmark Repo: https://github.com/liquidos-ai/autoagents-bench
I'd love your feedback on AutoAgents. We're determined to build the best agent framework with performance and safety-first design.
Thanks