Best AI Models (November 2026): Leaderboard Rankings, Prices and Which to Use

01 · Rankings

The leaderboards right now

Two different ways of measuring, and they agree on the top tier.

What is the best AI model in 2026?

On raw capability, Gemini 4 Carbon. Google announced it on October 28, 2026, and it leads the Artificial Analysis Intelligence Index with 59 points at high effort and the Arena text leaderboard with a preliminary 1,528 rating. It is only available to a subset of cybersecurity teams for now, so among models you can actually use, Claude Opus 5.5 (58) and GPT-6 Astra (57) are the strongest, with Gemini 4 Argon close behind on the Arena.

Source: Artificial Analysis: LLM leaderboard (read Nov 3, 2026); Arena (formerly LMArena): Text leaderboard (updated Oct 27, 2026); Google: Gemini 4 Carbon, our next frontier model (Oct 28, 2026)

What does the Artificial Analysis leaderboard show?

Artificial Analysis runs the same set of evaluations on more than 250 models and combines them into an Intelligence Index. On November 3, 2026 the top entries were Gemini 4 Carbon (high) at 59, Claude Opus 5.5 (max) at 58, Gemini 4 Carbon (medium) at 58, Claude Opus 5.5 (xhigh) and GPT-6 Astra (max) at 57, then GPT-6 Astra (xhigh and high) and GPT-6.1 Sol (max) at 55, and Claude Sonnet 5.5 (max) at 54. Effort settings matter: the same model scores higher, and costs more, when allowed to think longer.

Source: Artificial Analysis: LLM leaderboard (read Nov 3, 2026)

What does the Arena (LMArena) leaderboard show?

The Arena, formerly LMArena and Chatbot Arena, ranks models by millions of blind human votes between two answers, more than 7.9 million so far. On its October 27, 2026 update the text top five were Gemini 4 Carbon (high, 1,528, preliminary), Gemini 4 Argon (high, 1,520), Claude Opus 5.5 (high, 1,509), Meta's Muse Wonder (max, 1,507) and Claude Fable 5.1 (max, 1,504). Arena reflects which answers people prefer, not just benchmark accuracy.

Source: Arena (formerly LMArena): Text leaderboard (updated Oct 27, 2026)

Why do the leaderboards disagree sometimes?

They measure different things. Artificial Analysis scores fixed benchmarks for reasoning, knowledge, coding and agents, so it rewards raw problem-solving. Arena scores human preference in open-ended chats, so tone, formatting and helpfulness count too. Anthropic's models place many entries in the Arena top ten, while OpenAI's GPT-6 models rank higher on Artificial Analysis than in Arena. Use both, then test on your own tasks.

Source: Arena (formerly LMArena): Text leaderboard (updated Oct 27, 2026); Artificial Analysis: LLM leaderboard (read Nov 3, 2026)

02 · By use case

Which model is best for what

The top score is rarely the right buy.

What is the best AI model for coding?

Claude Opus 5.5. On Artificial Analysis's Terminal-Bench 4.0 runs at default effort it scored 52.5% inside Claude Code against 48.0% for GPT-6.1 Sol in Codex, and many developers moved to it after its September 22, 2026 launch. GPT-6.1 Sol does the same tasks for roughly a sixth of the cost per task, so it is the better pick for high-volume automation. See our Codex vs Claude Code comparison.

Source: CatDoes: Codex vs Claude Code 2026, pricing, limits and speed; Anthropic: Claude Opus 5.5; Mac Install Guide: Claude Code or ChatGPT Codex?

What is the best AI model for writing?

Claude models are the usual choice for long-form writing and editing, and they dominate the Arena's human-preference rankings, where Claude Opus 5.5 and Fable 5.1 both sit in the top five. Gemini 4 Argon is a strong alternative with very long context, and GPT-6.1 Terra was released on October 26, 2026 with what OpenAI calls clearer, more natural writing.

Source: Arena (formerly LMArena): Text leaderboard (updated Oct 27, 2026); OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026)

What is the best value AI model?

GPT-6.1 Sol. It scores 55 on the Artificial Analysis index at max effort with a blended price of about $1.27 per million tokens, against about $10.04 for Claude Opus 5.5 at max effort. Its list price is $2 input and $10 output per million tokens. GPT-6.1 Terra, at $1 and $5, is OpenAI's newer mid-tier option that it says matches GPT-6 Astra at a tenth of the price.

Source: Artificial Analysis: LLM leaderboard (read Nov 3, 2026); OpenAI API: Pricing; OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026)

What is the cheapest good AI model?

GPT-6 Luna and Claude Haiku 5.5 both cost $0.10 per million input tokens and $0.50 per million output tokens, and are fine for classification, extraction, summaries and chat at scale. Z.ai's GLM-5.3-Flash costs $0.15 and $0.50 and also reads images. Note that Haiku 5.5's price rises to five times its base rate once a prompt passes 100,000 tokens.

Source: CatDoes: Codex vs Claude Code 2026, pricing, limits and speed; Second Talent: Every GLM (Z.ai) model explained and compared (Sep 15, 2026)

03 · Prices

What the top models cost

API list prices per million tokens, input then output.

How much do the top AI models cost in the API?

Per million tokens: GPT-6 Astra $10 input and $50 output; Claude Fable 5.1 $10 and $50; Claude Opus 5.5 $4 and $20; GPT-6.1 Sol $2 and $10; Claude Sonnet 5.5 $2 and $10; GLM-5.3 $1.40 and $4.40; GPT-6.1 Terra $1 and $5; GPT-6 Luna and Claude Haiku 5.5 $0.10 and $0.50. Gemini 4 Carbon will launch at an introductory $2 and $10, rising to $4 and $20.

Source: OpenAI API: Pricing; CatDoes: Codex vs Claude Code 2026, pricing, limits and speed; Second Talent: Every GLM (Z.ai) model explained and compared (Sep 15, 2026); OpenAI: Introducing GPT-6.1 Terra (Oct 26, 2026); 9to5Google: Google announces Gemini 4 Carbon (Oct 28, 2026)

Do I need the API, or is a subscription enough?

For chatting, writing and occasional coding, a $20 subscription (ChatGPT Plus, Claude Pro or Google AI Pro at $19.99) gives you the top models with generous limits. Pay per token through the API when you are building an app, running automation or processing large volumes. Our LLM pricing guides compare both routes.

Source: GPTPrompts.AI: ChatGPT pricing 2026; Claude: Plans and pricing; Google One: Google AI plans

04 · Open models

The best open-weights models

Models you can download and run yourself.

What is the best open-source AI model in 2026?

On the Artificial Analysis index, DeepSeek V4.1 Flash is the top open-weights model in November 2026. Z.ai's GLM-5.3 scored 44.9 in September, then the highest of any open model, and GLM-5.3-Flash and GLM-5.2 are MIT licensed. Meta's Muse Glimmer 30B, released under Apache 2.0, runs on a single consumer GPU.

Source: Artificial Analysis: LLM leaderboard (read Nov 3, 2026); Second Talent: Every GLM (Z.ai) model explained and compared (Sep 15, 2026); Ollama Blog: Muse Glimmer from Meta Superintelligence Labs (Aug 10, 2026)

Can I run a top AI model on my own computer?

Not the frontier ones, which are closed or need data-centre hardware. Smaller open models such as Muse Glimmer 30B, Qwen and Gemma variants and quantized GLM-5.3-Flash builds run well on a recent Mac or a PC with a strong GPU through Ollama or LM Studio. See our LM Studio vs Ollama comparison to choose a runner.

Source: Ollama Blog: Muse Glimmer from Meta Superintelligence Labs (Aug 10, 2026); Kunal Ganglani: LM Studio vs Ollama, headless mode, pricing and speed

Sources

Read and checked October 2026. Policies change, so confirm anything time-sensitive on the source page.

The Best AI Models of 2026, Ranked