Skip to content

benchmark-report.md

Latest commit

 

History

History
48 lines (34 loc) · 1.52 KB

benchmark-report.md

File metadata and controls

48 lines (34 loc) · 1.52 KB

Benchmark Report (Local Machine)

Docs Home | Benchmarks | Framework Comparison | Ecosystem | Agent-Native

This report is generated from this repository with:

bun run bench:report

Machine

  • Date (UTC): 2026-02-18T06:24:49.777Z
  • Runtime: Bun 1.3.6
  • OS: linux 6.17.0-14-generic (x64)
  • CPU: Intel(R) Core(TM) i3-8130U CPU @ 2.20GHz
  • Logical cores: 4

Agent Overhead

Scenario ops/sec avg ms p99 ms vs direct
direct: processText() 526 1.899 6.880 1.00x
agent bridge: qirrel.parse_text 544 1.838 5.042 0.97x
mcp handler: tools/call 441 2.269 5.681 1.19x

Framework Comparison

Scenario ops/sec avg ms p99 ms vs direct
Direct handler 672637 0.001 0.009 1.00x
Qirrel AgentBridge 173444 0.006 0.019 3.88x
Qirrel MCP Handler 168128 0.006 0.016 4.00x
LangChain tool() 23150 0.043 0.112 29.06x
AI SDK tool() 476938 0.002 0.009 1.41x

Highlights

  • Qirrel AgentBridge is 7.49x faster than LangChain tool() in this run.
  • Qirrel MCP handler is 7.26x faster than LangChain tool() in this run.
  • Qirrel MCP handler runs at 0.35x of AI SDK tool() throughput while adding MCP compatibility.

Reproduce

  1. Close heavy background apps.
  2. Run bun run bench:report twice.
  3. Commit this file if the numbers are representative.