Epoch's capability index: Claude Opus 5.5 edges GPT-6 Astra for #1, Sonnet 5.5 level with Fable 5.1

Claude Opus 5.5 now tops the Epoch Capabilities Index, Epoch AI's single score built from dozens of benchmarks, at 167.35. GPT-6 Astra is a hair behind at 166.51, then Claude Sonnet 5.5 at 165.20 and Claude Fable 5.1 at 164.82.

THE TOP FOUR

From Epoch's own data file, read at 16:30 UTC on October 1, out of 270 scored models: 1. Claude Opus 5.5: 167.35 (confidence range 164.0 to 172.0) 2. GPT-6 Astra: 166.51 (163.1 to 171.1) 3. Claude Sonnet 5.5: 165.20 (161.8 to 169.4) 4. Claude Fable 5.1: 164.82 (161.8 to 169.1)

WHY "NARROWLY" MATTERS

The four ranges overlap almost completely. Opus 5.5 and Astra are less than a point apart with three-to-four-point uncertainty either side, so read the top of this board as a four-way tie with Opus 5.5 nominally ahead, not a clear win.

THE SONNET STORY

The one that stands out is Sonnet 5.5. Anthropic's mid-size model, at $2 in and $10 out per million tokens, scores level with Fable 5.1, which costs $10 and $50. Epoch puts it that way too: Sonnet 5.5 "has roughly matched" Fable 5.1.

WHAT IS NOT ON THE BOARD

GPT-6.1 Sol, released September 29, and Gemini 4 Argon, announced September 30 for trusted testers only, have no ECI score yet.