JIT-Agent Paper: Dynamic Harnesses Let Small Models Beat Giants While Cutting Costs by 36%
rohanpaul_ai · x · 2026-09-02
A new paper titled "JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness Evolution" proposes generating task-specific agent harnesses (memory, planning, tools) on demand, enabling smaller models to outperform stronger general models.
Key Findings:
- Performance Boost: DeepSeek-V4-Flash with JIT-Agent scored 85.1 on DeepSearchQA, beating GPT-5.6 (76.0). Across 18 pairs, JIT improved every backbone, with average gains of 7.7 points for GLM-5.2 and 8.8 for DeepSeek-V4-Flash.
- Cost Reduction: This wasn't just more inference. Compared to fixed harnesses like Claude Code and Codex, JIT-Agent achieved the lowest token usage and API cost in all 6 settings, averaging 36.0% lower cost than the cheapest fixed alternative.
More from coding & agent
- Open Source SEO Skill CLI: Integrates 70+ Audit Tools with MCP Server — iannuttall · 2026-09-02
- Opencode vs Deepseek harness: weak starts but relentless execution compared — GodComplecs · 2026-09-02
- Why retracted facts still poison your Agent's memory — go_kul_07 · 2026-09-02
- NextAdmin Launches: Open-Source Component Kit for Consistent AI Agents — Scobleizer · 2026-09-02
- A Year Later, LocalAI's Issue Tracker Is Run Mostly by Agents — IngwiePhoenix · 2026-09-02
- METATRON: Local LLM-powered automated penetration testing assistant — tom_doerr · 2026-09-02