China Telecom's TeleAgent desktop agent tested: long-task delivery, memory, and cost control
APPSO · wechat · 2026-09-09
APPSO tested TeleAgent, a desktop office agent from China Telecom built on its in-house Xingchen LLM and domestic compute stack. Across three scenarios—repeatedly revised event planning, student semester scheduling, and a five-person product launch—the agent handled multi-format inputs, tracked conflicting or unconfirmed dates instead of fabricating them, and updated deliverables consistently through rounds of requirement changes, aided by 400K+ token context and version retention. It supports shareable Skills, a ModelRouter that picks models by task complexity to cut credit costs (3000 free credits on signup), and safety guardrails like workspace limits and confirmation before critical file operations. The review frames long-task delivery quality, cost, and safety as the next battleground for office agents.
More from Apps
- Perplexity Search API lands in Hermes Agent with a 450B+ URL index — NousResearch · 2026-09-09
- Early users praise Muse agent's usefulness, naming their agents — infoxiao · 2026-09-09
- Muse hands-on: rich Connectors and goal tracking could make it the first mainstream consumer agent — omooretweets · 2026-09-09
- One person built a full media company from six Hermes bots — Roger_M_Taylor · 2026-09-09
- Deep-ML now covers a classic ML textbook: from-scratch NumPy to PyTorch projects — sharpeye_wnl · 2026-09-09
- An AR floor-sweeping app shipped and made money while your multiagent SaaS is still in dev — _jaydeepkarale · 2026-09-09