Unreal Agent
Unreal Labs released Unreal Agent, an asynchronous tool harness claiming up to 40% lower costs than Codex.
Unreal Labs published Unreal Agent, a Go SDK and runner that records tool calls as in-progress and finishes them in the background so the model avoids polling and heartbeats. The vendor reports up to 40% lower cost than Codex and 20% lower than Pi, with similar pass rates on Terminal-Bench 4.0, SWE-Atlas, DeepSWE 1.1, and Agents' Last Exam using GPT-6 Astra xhigh. The harness uses a small prompt, token-optimized tool results, and no sub-agents, and it accepts user steering while tools are still running.
- Asynchronous tool log lets the model schedule more work per turn.
- Terminal-Bench 4.0: 57.9% pass at $1,428 versus Codex at $2,350.
- Offers a Go library, runner executable, and Harbor-compatible benchmark runner.
- Prefers sandbox and token limits over harness approval hooks.
- Products
- Unreal AgentCodexPi
- Organizations
- Unreal Labs
- AI models
- GPT-6 Astra
From the original article
4 min read at unreallabs.ai
If you’re interested in frontier cost-efficiency for your AI agents, we’d love to work together! Get in touch: contact@unreallabs. ai .
By trollied · unreallabs.ai
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.