Best LLM Models 2026 — Live Ranking
Large language models on this page are ordered by arena ELO score. That score is a relative measure: it says how often a model wins in head-to-head comparisons judged by real users, not how fast it runs or what it costs. In 2026 the frontier is crowded — Anthropic, OpenAI, Google, xAI, Meta and the strongest Chinese labs (DeepSeek, Zhipu, Alibaba) all sit within a few dozen ELO points of each other, which means the gap between #1 and #8 is often a single benchmark. Use the table to see the current order, then open a model page to see what that model is actually good at.
15 models · 15 with an arena ELO score · updated 2026-10-01
| # | Model | Company | ELO | Adoption | Status |
|---|---|---|---|---|---|
| 1 | Gemini 4 Argon Google · #1 in the text arena · 4,942 community votes | 1,533 | 4,942 | Independent | |
| 2 | Claude Opus 5.5 Anthropic · #2 in the text arena · 3,932 community votes | anthropic | 1,512 | 3,932 | Independent |
| 3 | Claude Fable 5.1 Anthropic · #3 in the text arena · 11,241 community votes | anthropic | 1,511 | 11,241 | Independent |
| 4 | Claude Opus 5 Anthropic · #4 in the text arena · 28,351 community votes | anthropic | 1,506 | 28,351 | Independent |
| 5 | Claude Opus 4.6 Anthropic · #5 in the text arena · 77,193 community votes | anthropic | 1,504 | 77,193 | Independent |
| 6 | Gemini 3.8 Flash Google · #6 in the text arena · 24,828 community votes | 1,496 | 24,828 | Independent | |
| 7 | Claude Fable 5 Anthropic · #7 in the text arena · 37,900 community votes | anthropic | 1,492 | 37,900 | Independent |
| 8 | Mimo V2.6 Pro Xiaomi · #8 in the text arena · 4,074 community votes | xiaomi | 1,490 | 4,074 | Independent |
| 9 | Claude Opus 4.7 Anthropic · #9 in the text arena · 64,607 community votes | anthropic | 1,490 | 64,607 | Independent |
| 10 | Muse Spark 1.3 Meta · #10 in the text arena · 11,698 community votes | meta | 1,490 | 11,698 | Independent |
| 11 | Gemini 3.7 Flash Google · #11 in the text arena · 20,851 community votes | 1,488 | 20,851 | Independent | |
| 12 | Muse Spark 1.2 Meta · #12 in the text arena · 3,833 community votes | meta | 1,484 | 3,833 | Independent |
| 13 | Gemini 3.5 Flash Google · #13 in the text arena · 47,981 community votes | 1,481 | 47,981 | Independent | |
| 14 | Qwen3.8 Alibaba · #14 in the text arena · 22,809 community votes | alibaba | 1,481 | 22,809 | Independent |
| 15 | Gemini 3.1 Pro Google · #15 in the text arena · 121,225 community votes | 1,480 | 121,225 | Independent |
What actually separates the top LLM models
Arena ELO tells you which model wins comparisons most often, but it does not tell you which model fits your job. Before you pick one, look at four things the score does not cover: the price per million tokens or per generation, the context window or maximum output length, whether the model is available through an API or only inside a consumer app, and whether it can be self-hosted when the data cannot leave your infrastructure.
A model ranked eighth that is five times cheaper than the leader is usually the better business decision once volume grows. That is why each model page below lists what the model is used for rather than only its position.
Models in this ranking
- Gemini 4 Argon — google1533
- Claude Opus 5.5 — anthropic1512
- Claude Fable 5.1 — anthropic1511
- Claude Opus 5 — anthropic1506
- Claude Opus 4.6 — anthropic1504
- Gemini 3.8 Flash — google1496
- Claude Fable 5 — anthropic1492
- Mimo V2.6 Pro — xiaomi1490
- Claude Opus 4.7 — anthropic1490
- Muse Spark 1.3 — meta1490
- Gemini 3.7 Flash — google1488
- Muse Spark 1.2 — meta1484
- Gemini 3.5 Flash — google1481
- Qwen3.8 — alibaba1481
- Gemini 3.1 Pro — google1480
Other rankings
Put this ranking on your own site
The AI Bubbles visualisation is free to embed. Copy one line of HTML and the bubbles appear on your blog or documentation, with a link back to the full ranking here.
Get the embed code →