The top three on each LMArena board. Open the tier list above to see every model we track.
Everyday chat
Best for everyday chat: Gemini 4 Carbon
#1
Gemini 4 Carbon
#2
Gemini 4 Argon
#3
Claude Opus 5.5
Anthropic
Blind votes in everyday chats, 28 Oct 2026.
Coding
Best for coding: Claude Opus 5.5
#1
Claude Opus 5.5
Anthropic
#2
GPT-6 Astra
OpenAI
#3
Claude Sonnet 5.5
Anthropic
Blind votes on coding and building apps, 28 Oct 2026.
Getting tasks done
Best for getting tasks done: Claude Opus 5.5
#1
Claude Opus 5.5
Anthropic
#2
Gemini 4 Carbon
#3
Claude Sonnet 5.5
Anthropic
How well it finishes multi-step tasks, 2 Nov 2026.
Images
Best for images: GPT Image 2.5
#1
GPT Image 2.5
OpenAI
#2
GPT Image 2.6
OpenAI
#3
GPT Image 2
OpenAI
Blind votes between two pictures, 9 Nov 2026.
Video
Best for video: Gemini Omni 1.2 Flash
#1
Gemini Omni 1.2 Flash
#2
FLUX 3 Video
Black Forest Labs
#3
Grok Imagine Video 1.5
xAI
Blind votes between two clips, 9 Nov 2026.
Who makes each model, which one to pick, and what the newest release changed.
Claude
The current line-up is the 5.5 generation. Opus 5.5 is the best model for coding and getting tasks done. Sonnet 5.5 is close behind at half the price. Haiku 5.5 is the newest and cheapest.
Best overall: Claude Opus 5.5 · Best value: Claude Sonnet 5.5 · Cheapest: Claude Haiku 5.5
GPT (ChatGPT)
GPT-6 Astra is OpenAI’s strongest coder, but people rank it lower for everyday chat. GPT-6.1 Sol is the newest and nearly as good at coding for a fifth of the price. ChatGPT picks between them for you.
Best for coding: GPT-6 Astra · Best value: GPT-6.1 Sol · Cheapest: GPT-6 Luna
Gemini
Gemini 4 Carbon is the #1 model for everyday chat right now. Gemini 4 Argon is #2 at a mid-range price. Gemini 3.8 Flash is the fast, cheap option.
Best for chat: Gemini 4 Carbon · Best value: Gemini 4 Argon · Cheapest: Gemini 3.8 Flash
Grok
Grok 4.7 is xAI’s newest model. It sits #21 for coding and #16 for getting tasks done, and isn’t on the chat board yet.
Newest: Grok 4.7
Meta Muse
Meta’s Muse Wonder is its strongest model and #8 for chat. Muse Spark 1.3 powers Meta AI and costs much less.
Best overall: Muse Wonder · Best value: Muse Spark 1.3
Qwen
Qwen3.8 Max is Alibaba’s flagship: #16 for chat and #13 for coding. Qwen3.8 Flash Next and Qwen3.8 27B are open and cheap to run.
Best overall: Qwen3.8 Max · Open: Qwen3.8 Flash Next
DeepSeek
DeepSeek V4.1 Flash is the best open model for coding on these boards, and one of the cheapest to run.
Best overall: DeepSeek V4.1 Flash
Kimi
Moonshot’s Kimi K3.1 Pro is its newest: #9 for coding and #13 for tasks. Kimi K3 is open, so you can download and run it yourself.
Best for coding: Kimi K3.1 Pro · Open: Kimi K3
GLM (Z.ai)
Z.ai’s GLM-5.3 is open and top 30 for chat, coding and tasks. GLM-5.3 Flash is nearly as good for chat at about a tenth of the price.
Best overall: GLM-5.3 · Best value: GLM-5.3 Flash
MiMo (Xiaomi)
Xiaomi’s MiMo-V2.6-Pro is the highest-rated Chinese model for everyday chat (#12), and it’s open. MiMo-V2.6-Flash codes better (#22) for a fraction of the price.
Best overall: MiMo-V2.6-Pro · Best value: MiMo-V2.6-Flash
More from China
More Chinese labs on the boards. Tencent’s Hy4 Preview and MiniMax’s M3.1 Flash are rated for coding and tasks, not chat yet. Baidu’s ERNIE 5.1 is #35 for chat. ByteDance’s Seed 2.1 Pro is only on the coding board so far.
Best for coding: Hy4 Preview
Which AI model is best right now?
It depends on the job. On LMArena’s blind votes, Gemini 4 Carbon is #1 for everyday chat, Claude Opus 5.5 is #1 for coding and for getting tasks done, GPT Image 2.5 leads for images and Gemini Omni 1.2 Flash leads for video.
Is Claude or GPT (ChatGPT) better?
For coding, Claude Opus 5.5 is #1 and OpenAI’s GPT-6 Astra is #2. For everyday chat, Claude Opus 5.5 is #3 while OpenAI’s best, GPT-6.1 Sol, is #62. OpenAI leads for images with GPT Image 2.5.
How are these AI models ranked?
The tiers come from LMArena, where people compare two anonymous answers and vote for the better one. Thousands of votes become a rating for each model. We don’t run our own model tests, and nobody can pay to be ranked.
What’s the difference between an AI model and an AI tool?
A model is the engine, like Claude Opus 5.5 or Gemini 4 Carbon. A tool is the app you actually use, like Claude, ChatGPT or Gemini, and many tools let you pick between several models. For the best apps, see the AI tools tier list.
How often is this updated?
We refresh it from LMArena’s latest boards. The date under each board shows when its votes were last pulled.
Rankings from LMArena (CC BY 4.0). Release dates, memory and prices from models.dev.


















Copilot
NotebookLM
Canva
Grok








