Epoch AI says Claude Fable 5.2 “sets a new record on the Epoch Capabilities Index (ECI) at 174”. “The previous high was GPT-6 Astra’s 169”, the research group posted on X at 21:22 UTC on 9 November, the day Anthropic launched the model. The score is a third-party evaluation, not one of Anthropic’s own benchmark claims.
Fable 5.2 scores 7.2 points above Claude Fable 5.1, released 69 days earlier, by our arithmetic from Epoch’s data. Epoch’s chart puts its reasoning-era frontier trend at 16 points a year, which works out to about 3.0 points over 69 days, so the step is about 2.4 times that pace, by our arithmetic. Epoch says the score “is above what the pre-Astra trend for reasoning models would have predicted”.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on why products built on the same model can differ. It teaches the general idea only and does not discuss Epoch’s index or Fable 5.2.
The key insight: Epoch AI scores Claude Fable 5.2 at 174.24 on its Epoch Capabilities Index, 5.08 points above GPT-6 Astra, by our arithmetic from Epoch’s data file. Fable 5.2 sits 7.2 points above Fable 5.1, released 69 days earlier; Epoch’s 16-points-a-year trend implies about 3.0 over that span, by our arithmetic.
Epoch’s post adds: “This is notable given recent discussion of AI contributions to AI R&D, and we hope to release more work in this area soon”. The ECI, its index page says, “combines scores from many different AI benchmarks into a single” general-capability scale.
Epoch attaches two caveats, in its words: “This is based on testing we did prior to launch using only our own benchmarks”. And: “The ECI is continually updated as more benchmark results become available, and may change in the future”.
Epoch Capabilities Index scores for the top eight models in Epoch AI’s data file, downloaded 9 November 2026, with Epoch’s 90% ranges. Data: Epoch AI (CC-BY), a third-party evaluation; Fable 5.2’s score is based on Epoch’s pre-launch testing and may change. The 5.1-point gap is our arithmetic.
What Was Posted on X
Claude Fable 5.2 sets a new record on the Epoch Capabilities Index (ECI) at 174. The previous high was GPT-6 Astraโs 169. Fable 5.2โs score is above what the pre-Astra trend for reasoning models would have predicted.
This is notable given recent discussion of AI contributions to AI R&D, and we hope to release more work in this area soon.
This is based on testing we did prior to launch using only our own benchmarks. The ECI is continually updated as more benchmark results become available, and may change in the future.
Epoch AI’s post at 21:22 UTC on 9 November: Claude Fable 5.2 sets a new record on its Epoch Capabilities Index at 174, above GPT-6 Astra’s 169. Epoch says the score is based on its pre-launch testing using only its own benchmarks and may change as more results come in.
Where the Leaderboard Stands
In Epoch’s data file, downloaded at about 22:20 UTC on 9 November, Fable 5.2 scores 174.24 and GPT-6 Astra 169.16, a gap of 5.08 points by our arithmetic. Behind Astra, GPT-6.1 Sol (168.87), Claude Opus 5.5 (168.67) and GPT-6.1 Terra (168.11) sit within 1.05 points of it.
Epoch gives each score a 90% range; its chart calls them “90% bootstrap CIs”. Fable 5.2’s runs from 170.12 to 178.79 and Astra’s from 166.19 to 172.51, so the two overlap between 170.12 and 172.51, by our arithmetic.
The numbers move as results arrive. On 8 November, as we reported, Epoch’s data had Fable 5.1 at 166.9; the 9 November file shows 167.04.
What the Score Rests On
Epoch’s benchmark file holds 10 results for Fable 5.2, against 24 for GPT-6 Astra and 22 for Fable 5.1, by our count. All ten are for the model’s top (“max”) setting.
The file records 100% for Fable 5.2 on both private FrontierMath sets, Tiers 1-3 and Tier 4, against 98.6% and 97.6% for Astra and 94.9% and 87.8% for Fable 5.1. Fable 5.2 is ahead of Astra on seven of the ten benchmarks, by our count; Astra is ahead on GPQA Diamond (98.3% to 95.7%), OTIS Mock AIME (100% to 99.7%) and SpireClimb (34.0% to 19.6%).
The Trend Line
Epoch’s chart labels the line “Reasoning-era frontier trend: +16 ECI/year (90% CI: 14 to 18)”, fitted up to Fable 5.1. Over the 69 days from Fable 5.1’s 1 September date to 9 November, that range implies 2.6 to 3.4 points, by our arithmetic; Fable 5.2’s step is 7.2.
The data file we downloaded gives no trend value for 9 November. Epoch’s chart shows the Fable 5.2 point above the dashed extension of the trend line.
As we read it, the number to watch is the gap, not the rank. Four models behind Fable 5.2 sit within 1.05 points of each other, and Fable 5.2 is 5.08 points clear on the point estimate. Epoch’s own 90% ranges still overlap with Astra’s, and the score rests on 10 benchmark results, against 24 for Astra.
Epoch AI on X, 9 November 2026
“This is based on testing we did prior to launch using only our own benchmarks. The ECI is continually updated as more benchmark results become available, and may change in the future”
Three Implications
FOR BUYERS OF MODELS Epoch’s 174 is a third-party score, separate from Anthropic’s own launch claims, and Epoch says it may change as results come in.
FOR THE LEADERBOARD GPT-6 Astra, GPT-6.1 Sol, Claude Opus 5.5 and GPT-6.1 Terra sit within 1.05 points of each other in Epoch’s file, by our arithmetic.
FOR THE TREND Epoch’s trend of 16 points a year (90% CI 14 to 18) implies 2.6 to 3.4 points over 69 days, by our arithmetic; Fable 5.2’s step over Fable 5.1 is 7.2.
The essay argues: “Model weights are increasingly fungible — Sonnet, GPT, Gemini, and Grok cluster within a few points on benchmarks that mattered last year.”
As we read it, the four models behind Fable 5.2 in Epoch’s file fit that picture of clustering, while Fable 5.2’s 5.08-point lead on the point estimate sits outside it, on pre-launch numbers. The essay predates Fable 5.2 and does not discuss these scores.
What Is Not Established
The 174 rests on Epoch’s pre-launch testing on its own benchmarks, 10 results in its file, and Epoch says it may change. Fable 5.2’s 90% range overlaps Astra’s. The file we downloaded has no row for Mythos 5.2. Epoch has not yet published the AI R&D work its post mentions.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Epoch AI scores Claude Fable 5.2 at 174 on its capabilities index, 5.1 points above GPT-6 Astra, on pre-launch testing it says may change. Over 69 days, Fable 5.2 moved 7.2 points past Fable 5.1, about 2.4 times the pace of Epoch’s 16-points-a-year trend, by our arithmetic.
A note on sourcing. Scores, ranges and benchmark results come from Epoch AI’s post of 9 November, its chart and its ECI data files, downloaded the same day; Epoch says the scores may change. Gaps, day counts, the trend arithmetic and benchmark counts are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
