AI model ranking 2026: intelligence index and price as of October

People search for an "AI model ranking 2026" to quickly figure out which model to use for their task. Below is a ranking of the models available through Zukko by the Artificial Analysis Intelligence Index, with per-token prices as of October 7, 2026 and the cost of a typical request volume. And, separately, what such a ranking won't tell you.

Where the numbers come from

In the Zukko catalog, models have an "Intelligence index" column. It's the Intelligence Index from Artificial Analysis — a composite score computed across a set of benchmarks. The higher the number, the better the model did on that test set on average.

Two caveats:

• It's one number for everything. The index mixes different task types and doesn't say how a model will behave on your data specifically.
• Values are as of October 7, 2026. The index and prices change over time, and for some models the catalog has no index data. We don't rank those models; they're listed separately as "no data".

AI model ranking 2026 by intelligence index

Zukko prices in USD per 1M tokens, as of October 7, 2026. The "1,000 requests" column is the cost of a thousand typical requests at 2,000 input and 500 output tokens. Formula: 1000 × (2000 × input + 500 × output) / 1,000,000, i.e. 2 × input + 0.5 × output.

Checking one row: Claude Opus 5.5 — 2 × 0.8 + 0.5 × 4 = 1.6 + 2 = $3.60 per thousand requests. Models with the same index share a rank; within a pair they're sorted by price.

Models without index data

These models are in the catalog, but there's no index value for them. Compare them by price and on your own tasks.

For the current list and prices, call GET https://api.zukko.pro/v1/models (see the model list docs↗) or open the Models page in the dashboard.

How to read the ranking

Three tiers

• 50 and above: Claude Opus 5.5, GPT-6 Astra, Claude Fable 5.1, GPT-6.1 Sol. Candidates for tasks where mistakes are expensive.
• 40–45: GLM 5.3, Qwen 3.8 Max, GLM 5.3 Flash, GPT-5.6 Terra, Gemini 3.8 Flash, Qwen 3.8 Flash.
• Below 40: DeepSeek V4.1 Flash, GPT-6 Luna, DeepSeek V4 Pro. Candidates for high-volume simple tasks where answers are easy to check.

A couple of points is no reason to choose blindly

39 vs 41 or 52 vs 53 is a small gap for a composite score. If prices differ many times over at the same time, testing on your own examples decides, not the position in the table.

Price doesn't follow the index

GPT-6 Astra and Claude Fable 5.1 sit side by side on the index (53), but for our profile a thousand requests cost $2.765 and $31.50 respectively. DeepSeek V4 Pro (36) costs more than DeepSeek V4.1 Flash (39). Look at both numbers.

Models nothing else beats

A model is worth considering if no other model in the table is both cheaper and at least as high on the index. For the "2,000 input, 500 output" profile, those are:

• DeepSeek V4.1 Flash — 39, $0.045;
• GLM 5.3 Flash — 42, $0.10;
• GLM 5.3 — 45, $0.25;
• GPT-6.1 Sol — 52, $1.20;
• GPT-6 Astra — 53, $2.765;
• Claude Opus 5.5 — 58, $3.60.

The list depends on your token profile and hard requirements. GLM 5.3 and GLM 5.3 Flash have a 128K context and don't accept images; if you need both, Qwen 3.8 Max (45, $0.295), "beaten" on price, is back in play. So filter models by requirements first, then compare numbers.

What the ranking won't tell you

• Whether a model fits your task. Extracting fields from an invoice and planning a database migration are different tasks, but the score is one.
• How a model handles your prompt and format. The index measures a model on someone else's tests; you have your own instructions, language and JSON requirements.
• What your real traffic costs. The table figures are an example for one profile. Long answers, conversation history and caching change the total. Prices for all models with sample calculations are in the overview AI API prices in October 2026↗.
• What will happen in a month. Models get updated, and provider prices change.

The index is a tool for a first cut, not a replacement for testing. How to set up a small test on 30–100 examples is covered in detail in the article intelligence index: how to choose↗.

How to compare candidates yourself

All models are called through one OpenAI-compatible endpoint; only model changes. This script sends the same question to several models and prints the answer and tokens:

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.zukko.pro/v1", api_key=os.environ["OPENAI_API_KEY"])

MODELS = ["deepseek-v4.1-flash", "glm-5.3", "gpt-6.1-sol", "claude-opus-5-5"]
PROMPT = "Put a real task from your work here"

for model in MODELS:
    resp = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": PROMPT}],
        max_tokens=800,
    )
    print(f"=== {model}: input {resp.usage.prompt_tokens}, output {resp.usage.completion_tokens}")
    print(resp.choices[0].message.content)

Run 10–20 of your own tasks this way, rate the answers without looking at the model name, and pick the cheapest model that clears your quality bar. Endpoint reference is in the chat/completions docs↗. One key covers all models for this comparison — you can create it in the dashboard↗.

FAQ

Which AI model is the strongest in 2026?

By intelligence index as of October 7, 2026, Claude Opus 5.5 leads among Zukko models with a score of 58. It's a composite score across a benchmark set; for your task a cheaper model may well be enough.

Why isn't Claude Sonnet 5 in the ranking?

The catalog has no intelligence index value for it. It's listed in the "no data" table: compare it by price and on your own examples.

How often does the ranking change?

The index and prices can change at any time: new model versions come out, and providers revise prices. We state the date; current values are in the model catalog in the dashboard.

What should I pick on a tight budget?

Start with the models nothing else beats on the index/price pair: DeepSeek V4.1 Flash, GLM 5.3 Flash, GLM 5.3. Test them on your tasks and move to more expensive ones only if quality falls short.