LithosAI Launches LithosBox Millisecond Agent Sandboxes; Hits 727 TPS on GLM 5.3 Flash
JiaZhihao · x · 2026-10-06
LithosAI has launched a public preview of LithosBox: ultra-fast versioned sandboxes that agents can snapshot, fork, and restore in milliseconds, with Git-like state control.
- Agents can spin up thousands of environments to explore multiple paths, rebuild the winning state, and discard the rest—all in milliseconds.
- The company also claims the fastest GLM 5.3 Flash inference per Artificial Analysis: 727 TPS with 3.9s E2E for 10K input, and 937 TPS / 5.9s for 100K long context.
- It already holds #1 speed spots for Kimi K3 and DeepSeek V4.1 Flash, and argues the real bottleneck is end-to-end latency across the whole agent loop, not just token output speed.
More from coding & agent
- Gemini 3 Flash burned $14 and failed a $1 agent task Pro finished in 85 steps — XIFAQ · 2026-10-06
- This builder spends 5x more on scraping APIs than AI models — data is the edge — EXM7777 · 2026-10-06
- Interfaze open-weights MoA model for OCR, speech and GUI grounding on one 80GB GPU — charles_irl · 2026-10-06
- 10-person DTC brand runs on 16 AI agents: 90 improvements in 15 weeks, now sells AI setup to other brands — jacob_posel · 2026-10-06
- Eidon: Open-Source Self-Hosted AI Platform Runs a Team of Agents on Your Own Server — Quack66 · 2026-10-06
- Context Language Models: Letting LLMs Edit Their Own Context Boosts Long Tasks and Cuts Compute — Combinatorilliance · 2026-10-06