Swarms Rust multi-agent framework launches: 6ms cold start, up to 440x faster than LangGraph
KyeGomezB · x · 2026-10-04
Swarms released swarms-rs 0.3.0, an enterprise-grade multi-agent orchestration framework written in Rust, installable via cargo add swarms-rs.
Benchmarks against Swarms Python, LangGraph, and CrewAI claim:
- 6ms cold start — 130x to 440x faster than alternatives
- 3.7MB startup memory — 25x to 68x less
- 0.11ms framework overhead per LLM call — 10x to 88x less
- 100 parallel agents complete in 0.52s vs. an ideal 0.50s
The framework supports tools and MCP, runs any model through a single OpenRouter key (Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, xAI), and swapping models is a one-string change. Agents compose into sequential, concurrent, graph, and router workflows. The post includes a three-step quickstart: set up the project, build a first agent, and chain agents into a multi-model pipeline.
More from coding & agent
- Dev cuts CI costs to 1/24 of GitHub Actions by repurposing Celesto sandboxes as runners — aniketmaurya · 2026-10-04
- Open-source split-orchestrator plugin: Opus plans and reviews, Sol subagents implement — kevinkern · 2026-10-04
- Developer reflects on AI bots: stop being the bottleneck, stop managing sessions like files — alexcovo_eth · 2026-10-04
- Orchestrating local 27B subagents from Claude Code: three remote-control limitations — Distinct-Pie2389 · 2026-10-04
- Agents on real business workflows: the bottleneck is missing context, not the model — Yuanliu_AIGY · 2026-10-04
- What actually broke when I gave my agent full autonomy for a week — ladyshrekk · 2026-10-04