OpenClaw 2.0 vs Hermes: How 800K Extra Tokens Bought Better Verification
rohanpaul_ai · x · 2026-09-01
An experiment comparing OpenClaw 2.0 and Hermes Agent reveals significant differences in token usage despite similar costs for a video generation task. OpenClaw used 2.1M tokens ($4.5) with 10 self-fixes, while Hermes used 2.9M tokens ($4) with 20 self-fixes.
The key difference lies in verification:
- OpenClaw 2.0: Screenshotted frames at timestamps, counted pixels, and corrected visual errors like an inverted coat-tail direction.
- Hermes Agent: Built a custom render harness, measuring frame times, checking loop seams, and auditing scenes against the spec before human review.
The extra 800K tokens were spent on this double-checking. This run highlights that agent tokens aren't just inference costs; they are often the price of reliability.
Related event: OpenClaw 2.0 vs Hermes: Agent Differences Driven by Self-Checking(2 posts)→
More from coding & agent
- LLM one-shots a Minecraft 3x3 piston door by building its own Linux test infra — Vjeux · 2026-10-18
- Dev builds Mumbai-inspired 3D manga background generator from OpenStreetMap data — CurieuxExplorer · 2026-10-18
- Solo founder builds a money-printing data service on Claude's Pi Durable harness — MicahBerkley · 2026-10-18
- Dev builds Arch-based OS with custom desktop and browser using Claude Code — Snoo-10493 · 2026-10-18
- Five SaaS agent workflows fail the same way — and it's never model intelligence — usehive · 2026-10-18
- Claude skill generates hand-drawn interactive diagrams that teach kids how things work — CatAstro_Piyush · 2026-10-18