Top AI Text Models Arena Rankings | Generated by AI

Home 2026.04 English 中文

---

Question: What are the latest rankings on the Arena AI text leaderboard, and which models are currently at the top?

Answer:

As of April 2026, the Arena AI (formerly LMSys Chatbot Arena) text leaderboard is still one of the most widely watched rankings, based on real user preference voting (Elo system) rather than static benchmarks. (Medium)

🏆 Current Top (Text Arena – recent snapshot)

From the latest available leaderboard slices (especially math / overall categories):

1. Claude Opus 4.6 (thinking) — by Anthropic

   • Rank: #1
   • Elo: ~1516

2. GPT-5.4-high — by OpenAI

   • Rank: #2

👉 These two are consistently at the very top across multiple sub-leaderboards (math, reasoning, overall). (Arena)

---

🧠 Tier-1 Cluster (very close behind)

The leaderboard is extremely tight at the top, often within a few Elo points:

• Google DeepMind (Gemini 3 series, variants)
• xAI (Grok models)
• DeepSeek (V3.x / reasoning variants)
• Zhipu AI (GLM-5 family)
• Alibaba (Qwen 3.x large models)

For example, in open-model rankings (Feb 2026):

• GLM-5 ≈ 1455
• Qwen-3.5 ≈ 1454
• Kimi-K2.5 ≈ 1452 👉 basically tied within ~3 points (LinkedIn)

---

📊 Where Kimi stands (your earlier question)

• Moonshot AI (Kimi series)

• Kimi K2.x / K2.5:

  • Top 10 globally
  • Strong especially among open / semi-open models
  • But not #1 overall (still behind Claude / GPT tier)

It’s competitive, but the absolute top is still dominated by Claude + GPT.

---

⚠️ Important insight (many people miss this)

• Arena rankings are dynamic and noisy
• Small Elo gaps (5–20 points) are not meaningful differences

• Models are often A/B tested anonymously, so:

  • Prompt style can change results
  • Updates can reshuffle rankings quickly

---

🧭 Simple takeaway

• #1 right now: Claude Opus 4.x (thinking variants)
• #2: GPT-5.x high-end versions
• Top cluster: Gemini, Grok, DeepSeek, Qwen, GLM
• Kimi: strong, but just below top-tier frontier

---

If you want, I can break it down by coding / reasoning / creative writing, where rankings actually differ quite a bit.

References:

• https://arena.ai/leaderboard/text
• https://arena.ai/leaderboard/text/math
• https://medium.com/codex/how-the-arena-ai-leaderboard-helps-developers-pick-the-right-ai-model-ab87bcb5ed4a
• https://www.linkedin.com/posts/arenaai_top-10-open-models-february-2026-in-text-activity-7433282604247961600-xYiF

---

Back Donate