Skip to content
Model Index
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
LabsMethodology

Snapshot · Oct 8, 2026

Which SI model is actually worth using?

The current lineup from every major lab, measured on independent benchmarks — with real prices, speed and what each task costs to run. Filter it, sort it, compare models side by side, or let the finder narrow it down.

Explore all modelsHelp me choose
41
current models
19
labs
23
third-party benchmarks

Intelligence Index · top 5

Artificial Analysis
  1. 1
    Claude Opus 5.5

    Anthropic · Max effort

    57.6
  2. 2
    Claude Sonnet 5.5

    Anthropic · Max effort

    56.0
  3. 3
    Claude Fable 5.1

    Anthropic · Max effort

    53.4
  4. 4
    GPT-6 Astra

    OpenAI · Max effort

    52.7
  5. 5
    Gemini 4 ArgonLimited

    Google DeepMind · High effort

    52.6
Full ranking and what it measures

Who leads right now

The top model on each question people most often ask. Every figure comes from an independent evaluator, not from the lab.

  • Smartest overall

    Claude Opus 5.5

    57.6Intelligence Index

    Then Claude Sonnet 5.5 (56.0)

  • Best at coding

    Claude Opus 5.5

    97Coding score

    Then Claude Sonnet 5.5 (96)

  • Best at agentic work

    Claude Fable 5.1

    94Agents score

    Then Claude Opus 5.5 (91)

  • People's favourite

    Gemini 4 Argon

    1525Arena Text rating

    Then Claude Opus 5.5 (1504)

  • Most trustworthy facts

    Claude Opus 5.5

    46Omniscience index

    Then Claude Fable 5.1 (43)

  • Near the top for less

    GPT-6.1 Sol

    $0.72per task, within 15% of the top index

    Then Gemini 4 Argon ($1.99)

  • Fastest capable model

    Claude Haiku 5.5

    242tokens / second, index 40+

    Then Muse Spark 1.3 (209)

  • Best open weights

    MiMo-V2.6-Pro

    46.3Intelligence Index

    Then GLM-5.3 (44.8)

Intelligence for the money

Each dot is a model at its strongest setting. Up is smarter, left is cheaper. Models on the blue line are the efficient frontier: nothing else is both smarter and cheaper.

View as table
Intelligence IndexEfficient frontierOther models

The efficient frontier, cheapest first

  1. 1Mistral Small 411.3 · $0.02
  2. 2MiMo-V2.6-Flash37.9 · $0.06
  3. 3GPT-6 Luna38.1 · $0.07
  4. 4MiMo-V2.6-Pro46.3 · $0.13
  5. 5GPT-6.1 Sol51.8 · $0.72
  6. 6Gemini 4 Argon52.6 · $1.99
  7. 7GPT-6 Astra52.7 · $3.26
  8. 8Claude Sonnet 5.556.0 · $5.46
  9. 9Claude Opus 5.557.6 · $5.98

Cost per task is what Artificial Analysis paid to run its Intelligence Index, divided by the number of tasks. It includes reasoning tokens, so it reflects real spend better than list prices do. View as table

Latest releases

All by date
  • Claude Haiku 5.5

    Anthropic · Small & fast

    Oct 7
  • Mistral Large 4

    Mistral AI · Flagship

    Oct 6
  • Gemini 4 Argon

    Google DeepMind · Flagship

    LimitedSep 30
  • GPT-6.1 Sol

    OpenAI · Mid-tier

    Sep 29
  • Claude Sonnet 5.5

    Anthropic · Mid-tier

    Sep 28
  • Claude Opus 5.5

    Anthropic · Flagship

    Sep 22
  • GPT-6 Luna

    OpenAI · Small & fast

    Sep 22
  • MiMo-V2.6-Flash

    Xiaomi · Small & fast

    Sep 22

Rank by what you need

  • Overall

    Composite indexes and broad human preference — the quickest read on general capability.

    Leader

    Claude Opus 5.5

  • Coding

    Writing, running and shipping software: terminal agents, scientific code, app building and migrations.

    Leader

    Claude Opus 5.5

  • Agents & knowledge work

    Multi-step professional tasks with tools and files — finance, legal, tax and office deliverables.

    Leader

    Claude Fable 5.1

  • Reasoning & science

    Hard problems with no shortcut: research-level science, novel puzzles and abstract reasoning.

    Leader

    Claude Opus 5.5

  • Knowledge & reliability

    Does it know the answer, admit when it does not, and keep track across very long inputs?

    Leader

    Claude Opus 5.5

  • Vision

    Understanding images, charts, screenshots and documents.

    Leader

    Claude Opus 5.5

Browse by lab

All labs
  • AnthropicClaude · 4
  • OpenAIGPT · 5
  • Google DeepMindGemini · 5
  • MetaMuse · 2
  • SpaceXAIGrok · 1
  • XiaomiMiMo · 2
  • AlibabaQwen · 3
  • Z.aiGLM · 2
  • StepFunStep · 1
  • Moonshot AIKimi · 2
  • DeepSeekDeepSeek · 2
  • Mistral AIMistral · 3
  • MiniMaxMiniMax · 1
  • TMThinking Machines LabInkling · 2
  • TencentHy · 2
  • NVIDIANemotron · 1
  • CohereCommand · 1
  • AmazonNova · 1
  • MicrosoftMAI · 1
Model Index

An independent snapshot of frontier SI models and how they perform on third-party benchmarks.

Data as of Oct 8, 2026

Explore

  • All models
  • Compare
  • Benchmarks
  • Model finder
  • Labs
  • Methodology & sources

Sources

Benchmark results come from Artificial Analysis, Arena, Vals AI and ARC Prize. Specifications and prices come from each lab's own documentation.

Not affiliated with any lab or evaluator. Names and logos are trademarks of their respective owners and are used only to identify them. Scores change often — check the linked source before making a decision that depends on one.

  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder