Phone Use Arena launches: 17 phone-use agents ranked, task costs range $0.008-$1.23
ycombinator · x · 2026-11-04
Coarena launched the Phone Use Arena, a blind-evaluated leaderboard for phone-operating agents.
- 17+ models go head-to-head, judged by blind human and model raters, spanning OpenAI, Anthropic, Kimi, Meta, Google and more
- 74 of 101 metrics can be measured on a phone; where a phone can't measure something, the board explains why
- Cost per finished task ranges from $0.008 to $1.23, a 150x spread
- GPT-6.1 Terra leads overall (1116±44), followed by Kimi K3 (1100) and Kimi K3.1 Pro (1067); open models like Qwen3.8-2.4T, GLM 5.3 Flash and MiMo V2.6 Pro also rank
- Updates every 10 seconds with per-use-case rankings (settings, clock/calendar, mobile web) and CSV export
More from coding & agent
- Xogot ships built-in Pi.dev-powered coding agent for Godot on iPad — mitsuhiko · 2026-11-04
- DeepSeek quietly launches "DeepSeek Harness" YouTube channel, hinting at plugin-based agent framework — gnukeith · 2026-11-04
- Sharing agent sessions across a team breaks knowledge silos — and AI can summarize your teammates' workflows — steipete · 2026-11-04
- Apgar: open-source release health checker catches agent regressions without an LLM judge — krstCB · 2026-11-04
- Claude-based agent transcribes entire site with images into context for Claude 3 Opus — repligate · 2026-11-04
- Free no-sign-in MCP server compiles LaTeX (pdflatex/xelatex) and explains errors in plain words — PurplelinkPL · 2026-11-04