Skip to main content
idapt
AccueilModèles IATarifs
Sign inStart free
All terms

Learn

What is the Epoch Capabilities Index (ECI)?

Last reviewed July 16, 2026

The Epoch Capabilities Index (ECI) is a single capability score fit across many models and 50+ benchmarks using item response theory: benchmarks differ in difficulty, models differ in ability, and the statistical fit estimates both at once. Published openly by Epoch AI (data under CC-BY), it makes models comparable on one scale even when no two ran identical test sets, and it carries confidence bounds, so uncertainty is visible instead of hidden.

How an IRT fit works

Item response theory, borrowed from standardized testing, models the probability that a model of a given ability passes a benchmark of a given difficulty. Feeding it many models' results on 50+ overlapping benchmarks yields difficulty estimates for the tests and ability estimates for the models simultaneously. New results refine the whole fit, and models that never ran the same benchmark still land on one scale.

What the index cannot tell you

A composite capability score hides shape: two models with equal ECI can differ widely on coding versus math versus factuality, and the index says nothing about price, speed, or context window. Confidence bounds matter too: sparse evaluations mean wide intervals. Use ECI to shortlist, then check the named benchmarks and live numbers closest to your work.

Top models on Epoch Capabilities Index

Full leaderboard
  1. 1GPT 5.5 Pro91.8
  2. 2Claude Fable 591.6
  3. 3GPT 5.589.6
  4. 4Claude Opus 4.889.2
  5. 5GPT 5.4 Pro89.1
  6. 6Grok 4.588.5
  7. 7Claude Opus 4.787.4
  8. 8GPT 5.487.4
  9. 9GPT 5.3 Codex87.0
  10. 10Claude Opus 4.686.8

Peak scores from the compiled benchmark data; reasoning models use their highest effort tier.

In idapt

idapt's capability rankings and model pages use ECI rescaled to 0-100, re-fit for catalog models Epoch has not scored, always shown with confidence bounds and a provisional marker where the interval is wide, and never blended with price.

Capability rankingsHow we rank models

Frequently asked

Related terms

GPQAHumanity's Last ExamMETR time horizonReasoning effort

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

AI Gateway

OpenAI-, Anthropic-, and OpenRouter-compatible endpoints with routing you control.

See it in practice

The fastest way to understand it is to use it. Start free, no credit card.

Try idapt free
  • Accueil
  • Tarifs
  • Modèles IA
  • Modèles d'image
  • Modèles de voix
  • Modèles vidéo
  • Classements
  • New models
  • Model status
  • Multi-Model Chat
  • Voice Mode
  • Voice HUD
  • Web Search
  • Image Generation
  • Video Generation
  • Audio Generation
  • Transcription
  • Drive
  • Secrets
  • Sharing
  • Workspaces
  • Tasks
  • Memory
  • Agents
  • Subagents
  • Automations
  • Skills
  • Code Execution
  • Computers
  • Computer Use
  • Cloud Computers
  • Local AI
  • AI Gateway
  • API & SDK
  • CLI
  • MCP
  • Tunnels
  • All features →
  • Blog
  • Cas d'usage
  • Comparisons
  • Best of
  • Learn
  • Free tools
  • Prompts
  • Templates
  • Changelog
  • Centre d'aide
  • FAQ
  • Confidentialité
  • Comparer tous les modèles
  • Support
  • Développeurs
  • Quickstarts
  • Référence API
  • Tarifs API
  • CLI
  • MCP
  • Downloads
  • Desktop
  • Badges and embeds
© idapt[email protected]ConditionsPolitique de confidentialitéMentions légalesSignaler un contenu
X (Twitter)