Live ranking, refreshed every 2 days
LLM Leaderboard: Best AI Models Ranked
As of October 8, 2026, the top-ranked model on LMArena's text leaderboard is gemini-4-argon-high by Google with a score of 1,525, 18 points ahead of claude-opus-5.5-high by Anthropic.
Updated October 13, 2026. Showing the last saved copy of the leaderboard, dated October 8, 2026, because the live source could not be reached.
Rankings come from blind head-to-head votes on LMArena, the largest public human-preference arena for AI models. We read its open dataset every 2 days and compute who leads, by how much, and how that has changed.
Best AI chatbot: which assistant ranks first when blind-vote quality is combined with real usage.
Ranking: top 10 models
Scores come from blind head-to-head votes on LMArena's text leaderboard. Rank is LMArena's published rank, which accounts for confidence intervals, so several models can share a rank. The top 10 are shown below, and ranks 11 to 60 follow in the expandable list.
| Rank | Model | Lab | Score | Votes | License |
|---|---|---|---|---|---|
| 1 | gemini-4-argon-high | 1,525 ±9 | 4,892 | Proprietary | |
| 2 | claude-opus-5.5-high | Anthropic | 1,507 ±8 | 6,272 | Proprietary |
| 3 | claude-opus-4-6-high | Anthropic | 1,504 ±3 | 79,415 | Proprietary |
| 4 | claude-fable-5-high | Anthropic | 1,504 ±4 | 40,026 | Proprietary |
| 5 | claude-opus-4-7-high | Anthropic | 1,501 ±4 | 65,192 | Proprietary |
| 6 | claude-fable-5.1-max | Anthropic | 1,501 ±6 | 13,063 | Proprietary |
| 7 | claude-opus-4-6 | Anthropic | 1,498 ±3 | 84,095 | Proprietary |
| 8 | gemini-3.8-flash-high | 1,497 ±5 | 31,437 | Proprietary | |
| 9 | muse-spark-1.3-max | Meta | 1,494 ±6 | 14,694 | Proprietary |
| 10 | claude-opus-4-7 | Anthropic | 1,494 ±4 | 66,315 | Proprietary |
Show ranks 11 to 60
Rank Model Lab Score License 11 muse-spark-1.2 (xHigh) Meta 1,492 ±8 Proprietary 12 muse-spark-1.1 Meta 1,491 ±4 Proprietary 13 claude-opus-5-high Anthropic 1,490 ±4 Proprietary 14 claude-opus-5-max Anthropic 1,489 ±5 Proprietary 15 muse-spark Meta 1,489 ±6 Proprietary 16 kimi-k3-max Moonshot AI 1,488 ±5 Open weights 17 gemini-3.1-pro-preview Google 1,487 ±3 Proprietary 18 gemini-3.7-flash-high Google 1,486 ±5 Proprietary 19 gemini-3-pro Google 1,485 ±4 Proprietary 20 gpt-5.6-sol-xhigh OpenAI 1,485 ±4 Proprietary 21 gpt-6.1-sol-max OpenAI 1,484 ±9 Proprietary 22 qwen3.8-max Alibaba (Qwen) 1,483 ±5 Proprietary 23 gemini-3.6-flash-high Google 1,482 ±4 Proprietary 24 gpt-5.5-high OpenAI 1,482 ±4 Proprietary 25 claude-opus-4-8-high Anthropic 1,481 ±4 Proprietary 26 mimo-v2.6-pro Xiaomi 1,480 ±9 Open weights 27 glm-5.3-max Z.ai (Zhipu) 1,478 ±5 Open weights 28 gemini-3.5-flash-high Google 1,478 ±4 Proprietary 29 gpt-5.5 OpenAI 1,477 ±4 Proprietary 30 gemini-3.5-flash-medium Google 1,476 ±4 Proprietary 31 gpt-5.2-chat-latest-20260210 OpenAI 1,476 ±4 Proprietary 32 claude-sonnet-5.5-xhigh Anthropic 1,476 ±9 Proprietary 33 glm-5.2-max Z.ai (Zhipu) 1,475 ±4 Open weights 34 gpt-5.4-high OpenAI 1,475 ±4 Proprietary 35 gpt-6-astra-max OpenAI 1,475 ±7 Proprietary 36 qwen3.7-max-preview Alibaba (Qwen) 1,475 ±10 Proprietary 37 grok-4.20-beta1 xAI 1,475 ±5 Proprietary 38 glm-5.3-flash Z.ai (Zhipu) 1,475 ±5 Open weights 39 deepseek-v4.1-flash-max DeepSeek 1,475 ±7 Open weights 40 claude-opus-4-8 Anthropic 1,474 ±4 Proprietary 41 claude-opus-4-5-20251101-high-32k Anthropic 1,474 ±4 Proprietary 42 gpt-5.5-instant OpenAI 1,473 ±5 Proprietary 43 gemini-3-flash Google 1,473 ±4 Proprietary 44 claude-sonnet-4-6 Anthropic 1,472 ±4 Proprietary 45 grok-4.20-beta-0309-reasoning xAI 1,472 ±4 Proprietary 46 grok-4.20-multi-agent-beta-0309 xAI 1,470 ±4 Proprietary 47 claude-opus-4-5-20251101 Anthropic 1,470 ±3 Proprietary 48 mimo-v2.5-pro Xiaomi 1,468 ±4 Open weights 49 ernie-5.1 Baidu 1,467 ±5 Proprietary 50 gpt-5.6-terra-xhigh OpenAI 1,466 ±4 Proprietary 51 grok-4.5 xAI 1,466 ±4 Proprietary 52 grok-4.1-thinking xAI 1,465 ±3 Proprietary 53 gpt-5.4 OpenAI 1,465 ±4 Proprietary 54 qwen3.5-max-preview Alibaba (Qwen) 1,465 ±5 Proprietary 55 deepseek-v4-pro-high-20260813 DeepSeek 1,465 ±7 Open weights 56 glm-5.1 Z.ai (Zhipu) 1,464 ±4 Open weights 57 claude-sonnet-5-high Anthropic 1,461 ±4 Proprietary 58 kimi-k2.6 Moonshot AI 1,461 ±4 Open weights 59 qwen3.6-max-preview Alibaba (Qwen) 1,460 ±8 Proprietary 60 grok-4.1 xAI 1,459 ±3 Proprietary
Showing the top 60 of 414 models on this leaderboard. Data from October 8, 2026.
Best model by lab
Each lab's highest-ranked model on this leaderboard, and how far it sits from the top.
| Lab | Best model | Rank | Score | Behind #1 |
|---|---|---|---|---|
| gemini-4-argon-high | 1 | 1,525 | Leader | |
| Anthropic | claude-opus-5.5-high | 2 | 1,507 | 18 pts |
| Meta | muse-spark-1.3-max | 9 | 1,494 | 31 pts |
| Moonshot AI | kimi-k3-max | 16 | 1,488 | 37 pts |
| OpenAI | gpt-5.6-sol-xhigh | 20 | 1,485 | 40 pts |
| Alibaba (Qwen) | qwen3.8-max | 22 | 1,483 | 42 pts |
| Xiaomi | mimo-v2.6-pro | 26 | 1,480 | 45 pts |
| Z.ai (Zhipu) | glm-5.3-max | 27 | 1,478 | 47 pts |
| xAI | grok-4.20-beta1 | 37 | 1,475 | 51 pts |
| DeepSeek | deepseek-v4.1-flash-max | 39 | 1,475 | 51 pts |
| Baidu | ernie-5.1 | 49 | 1,467 | 58 pts |
| ByteDance | dola-seed-2.0-pro | 65 | 1,456 | 69 pts |
Which AI lab leads?
The lab with the most models at the top of the board is the one with the deepest bench, not only the single best model.
Anthropic holds 7 of the top 10 models (70%). Anthropic holds 9 of the top 20 models (45%).
Models in the top 10 by lab
Models in the top 20 by lab
Show data table
| Lab | Top 10 | Top 20 |
|---|---|---|
| Anthropic | 7 | 9 |
| 2 | 5 | |
| Meta | 1 | 4 |
| Moonshot AI | 0 | 1 |
| OpenAI | 0 | 1 |
Who has held the number one spot
Between August 28, 2024 and October 8, 2026 LMArena published 188 text leaderboards. The lab at #1 changed 7 times. Google has held #1 since September 30, 2026, 8 days as of the latest leaderboard. The current leader is gemini-4-argon-high.
- OpenAI
- Anthropic
| Lab | Days at #1 | Share of period | Longest run (days) |
|---|---|---|---|
| OpenAI | 314 | 41% | 300 |
| Anthropic | 240 | 31% | 236 |
| 217 | 28% | 122 |
Every reign at #1
| From | To | Lab | Model | Days |
|---|---|---|---|---|
| September 30, 2026 | October 8, 2026 | gemini-4-argon-high | 8 | |
| February 6, 2026 | September 30, 2026 | Anthropic | claude-opus-5.5-high | 236 |
| October 7, 2025 | February 6, 2026 | gemini-3-pro | 122 | |
| October 3, 2025 | October 7, 2025 | Anthropic | claude-sonnet-4-5-20250929-thinking-32k | 4 |
| August 21, 2025 | October 3, 2025 | gemini-2.5-pro | 43 | |
| August 7, 2025 | August 21, 2025 | OpenAI | gpt-5-high | 14 |
| June 24, 2025 | August 7, 2025 | gemini-2.5-pro | 44 | |
| August 28, 2024 | June 24, 2025 | OpenAI | o3-2025-04-16 | 300 |
History uses the style-controlled text leaderboard, which the dataset records from August 28, 2024. Durations run between published leaderboards, so they are approximate to the publication interval.
Top 10 share by lab over time
The share of the ten highest-scoring models each lab has held at every published leaderboard.
Data table: top 10 share by quarter
| Date | OpenAI | Anthropic | xAI | |
|---|---|---|---|---|
| September 27, 2024 | 50% | 10% | 10% | 10% |
| December 30, 2024 | 40% | 20% | 20% | 0% |
| March 30, 2025 | 50% | 20% | 0% | 10% |
| June 24, 2025 | 50% | 10% | 20% | 10% |
| September 30, 2025 | 50% | 20% | 10% | 10% |
| December 30, 2025 | 10% | 40% | 30% | 20% |
| March 31, 2026 | 20% | 30% | 30% | 20% |
| June 25, 2026 | 10% | 60% | 20% | 0% |
| September 30, 2026 | 0% | 70% | 10% | 0% |
| October 8, 2026 | 0% | 70% | 20% | 0% |
Quarterly sample plus the latest leaderboard. The chart uses all 188 leaderboards.
Open vs proprietary: how big is the gap?
The best open-weight model is kimi-k3-max by Moonshot AI at 1,488, 37 points behind the best proprietary model, gemini-4-argon-high by Google at 1,525. Their 95% confidence intervals do not overlap, so the gap is statistically meaningful.
15 proprietary models rank above the best open-weight model, and 1 of the top 20 is open-weight.
Open-weight here means any license other than Proprietary, including licenses with usage limits. It is a broader definition than the OSI one.
Other leaderboards
LMArena runs separate arenas for different jobs. The best model for one is often not the best for another.
Other arenas
- WebDev: claude-opus-5.5-max by Anthropic, 1,813
- Vision: claude-fable-5-high by Anthropic, 1,308
- Search: gpt-5.6-sol-xhigh by OpenAI, 1,256
- Document: claude-fable-5 by Anthropic, 1,499
- Text to image: gpt-image-2.5-sunburst by OpenAI, 1,425
- Image editing: gpt-image-2.5-sunburst by OpenAI, 1,524
- Text to video: gemini-omni-1.1-flash by Google, 1,516
Text arena by category
How to read this leaderboard
The text arena is LMArena's flagship: people send any prompt, compare two anonymous answers and vote. It measures overall preference across writing, reasoning, coding and everyday questions.
- Scores come from blind votes
- LMArena shows a person two anonymous answers to the same prompt and asks which is better. Those head-to-head votes are fitted with a Bradley-Terry model, the statistical successor to Elo, to give each model one arena score. On the Elo scale LMArena uses, a 100-point lead means the higher model is expected to win about 64% of matchups.
- Confidence intervals decide ties
- Every score carries a 95% confidence interval. When two intervals overlap, the order of those two models is not settled. LMArena's rank column already accounts for this, which is why more than one model can hold the same rank.
- Votes show how settled a score is
- Votes counts the head-to-head battles behind a score. A new model starts with a wide interval that narrows as votes accumulate, so a high score on few votes deserves caution.
- Style control removes the length bias
- The text leaderboard uses style control, LMArena's default since May 2025. It statistically adjusts for answer length and formatting, so a model cannot climb simply by writing longer or more heavily formatted answers.
- Why it differs from benchmarks
- Arena scores measure what people prefer on open-ended prompts. Fixed benchmarks measure accuracy on set questions. A model can top one and trail on the other, so the two answer different questions.
- Why it differs from usage share
- A high rank does not mean many people use a model. Usage depends on price, distribution and habit. For adoption rather than quality, see AI market share.
Methodology and source
All rankings are read from the LMArena leaderboard dataset on Hugging Face, published by LMArena (formerly Chatbot Arena) under the Creative Commons Attribution 4.0 license. Credit for the underlying votes, scores and confidence intervals belongs to LMArena. Essentras is not affiliated with LMArena.
This page reads the text_style_control subset, category overall, from the dataset's latest split. Scores, intervals, ranks, vote counts and license labels are shown as published. Essentras adds the grouping by lab and the derived figures on this page: gaps, lab shares, open vs proprietary and, on the main page, the history of who held #1.
Data refreshes automatically every 2 days. If the source cannot be reached, the page falls back to the last saved copy and says so at the top. Models are never merged or renamed; model and lab names appear exactly as LMArena publishes them.
Open-weight means any license other than Proprietary. History is limited to what the dataset records: the style-controlled text leaderboard from August 28, 2024. We do not scrape any other ranking site.
Track how AI models rank and describe your brand with our AI SEO service. Want these models to name your brand? Read our guide to LLM SEO.
Frequently asked questions
What is the best AI model right now?
As of October 8, 2026, gemini-4-argon-high by Google leads LMArena's text leaderboard with a score of 1,525. "Best" depends on the job: the coding, vision and search leaderboards often have different leaders, linked above.
What is the best LLM for coding?
On LMArena's WebDev arena, which tests building working web apps, claude-opus-5.5-max by Anthropic leads with 1,813 as of October 8, 2026. In the coding category of the text arena, gemini-4-argon-high by Google leads with 1,562.
Which is better, ChatGPT, Gemini or Claude?
On the text leaderboard as of October 8, 2026: Google's best model, gemini-4-argon-high, ranks 1 with 1,525, Anthropic's best model, claude-opus-5.5-high, ranks 2 with 1,507, and OpenAI's best model, gpt-5.6-sol-xhigh, ranks 20 with 1,485. Google is highest, 18 points ahead of the next of the three and outside its confidence interval.
What is an LLM leaderboard?
An LLM leaderboard ranks large language models by measured performance. Some use fixed benchmarks with right answers. LMArena's uses human preference: people compare two anonymous answers and vote, and the votes become a score for each model.
How are AI models ranked?
Each vote is a head-to-head comparison between two models. A Bradley-Terry model turns thousands of these into a score per model, with a 95% confidence interval. Models are ranked by score, and models whose intervals overlap can share a rank.
Is the best open-source model close to the best proprietary model?
The best open-weight model is kimi-k3-max by Moonshot AI at 1,488, 37 points behind the best proprietary model, gemini-4-argon-high by Google at 1,525. 15 proprietary models rank above the best open-weight model, and 1 of the top 20 is open-weight. Their 95% confidence intervals do not overlap, so the gap is statistically meaningful. "Open" here means any license other than Proprietary, which is broader than the OSI definition.
How often does the leaderboard change?
LMArena published 188 text leaderboards from August 28, 2024 to October 8, 2026, and the lab at #1 changed 7 times in that period. This page re-reads the dataset every 2 days and shows the date of the latest published data.
Models rank themselves. What do they say about you?
A leaderboard shows which AI models are strongest. It does not show what those models tell your customers about your company. Run the free check and see the answers.