Claude Opus 5.5 Tops Text Arena at 1509, a New Record for Text Models Claude Opus 5.5 has only been out for four days, and it already sits atop the text-model arena. On September 26, LMArena's official account on X published its latest rankings: Claude Opus 5.5 (High) debuted at No. 1 on Text Arena with 1509 points, the highest score ever recorded on that leaderboard.
Text Arena is LMArena's text-model leaderboard, ranked by anonymous head-to-head votes from users around the world across math, coding, creative writing, and more. Unlike one-shot static benchmarks, it captures something closer to "which model feels stronger in real use." The 1509 score (with a margin of about ±12) puts Opus 5.5 past the previous leader — just four days after its official launch on September 22.
The cheaper model took the crown
The interesting part is the price. Opus 5.5's API pricing is $4 per million input tokens and $20 per million output tokens, with performance close to the flagship Fable 5.1, a 1-million-token context window, and up to 128K tokens of output in a single run. By comparison, OpenAI's flagship GPT-6 Astra lists at $10 / $50. In other words, the No. 1 model on the board is also one of the cheapest at the frontier — "strongest" and "cheapest" rarely coincide, and this time is an exception. (For the full launch pricing and specs, see our earlier Claude Opus 5.5 launch breakdown.)
What the leaderboard proves — and what it doesn't
Still, take the No. 1 with a grain of salt. Text Arena is fundamentally a human-preference vote: it reflects overall perceived quality, not a hard metric for any single capability, and scores will keep shifting as more votes come in. For everyday users, the practical takeaway is about model selection: if you're picking a model for long coding sessions or agentic work, Opus 5.5 now carries both the "crowd-voted No. 1" and "one of the cheapest flagships" labels, so the cost of trying it is lower than ever. That doesn't make it the best at everything — rivals remain close on terminal-agent and math-reasoning benchmarks. Leaderboards change; the price war probably won't stop.
