Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
12:09 PM · Jun 12, 2026 · 130.8K Views
Anthropic's Claude Fable 5 scores 88% on FrontierMath Tier 4, outperforming OpenAI on the benchmark for the first time
Story Overview
Anthropic's Claude Fable 5, the consumer version of its Mythos-class model released around June 9-10, posted 87 percent on FrontierMath Tiers 1-3 and 88 percent on Tier 4. Epoch AI Research flagged the results, noting this is the first time an Anthropic model has led OpenAI on the Tier 4 problems, which are authored by expert mathematicians and kept unpublished.
1512K75236242.3K
Benchmark Leap
Rapid gains keep beating internal forecasts
The Tier 4 result cleared even bullish mid-year predictions by roughly 25 points, extending a streak of steady math-reasoning lifts across recent Claude releases.
ECI Signal
One number hints at broader capability
The FrontierMath performance implies an Epoch Capabilities Index of 164, though the exact composite figure for this model variant remains an inference rather than a separately published total.
Sentiment
Positive users praise Claude Fable 5's FrontierMath benchmark wins over OpenAI as evidence of fast reasoning gains, while negative users dismiss the results as irrelevant shilling or Anthropic hype that ignores real-world needs.
Pos
44.3%
Neg
55.7%
36 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS58.2KBOOKMARKS69LIKES518RETWEETS16
Ethan Mollick@emollick
The shape of the graph is getting very familiar.
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
3hViews 58.2KLikes 518Bookmarks 69
REPLIES22
Chubby♨️@kimmonismus
Looking at the graph, I think Fable 5 will only maintain its lead up to GPT-5.6.
And secondly, I think the benchmark will soon be completely saturated.
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
2hViews 17.3KLikes 302Bookmarks 28
Lisan al Gaib@scaling01
Anthropic is currently just scale-mogging everyone
scaling just works and will continue to work
Lisan al Gaib@scaling01
Fable 5 scored 88% on FrontierMath Tier 4 this implies an ECI score of 164
it's also the first time Anthropic has a model that is better at math than OpenAI
3hViews 14.5KLikes 239Bookmarks 21
Lisan al Gaib@scaling01
Fable 5 scored 88% on FrontierMath Tier 4 this implies an ECI score of 164
it's also the first time Anthropic has a model that is better at math than OpenAI
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
3hViews 23.4KLikes 206Bookmarks 24
Nat McAleese@__nmca__
while it is consistent with my predictions on paper, this still gives me an astonishing amount of future shock
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
2hViews 4KLikes 73Bookmarks 9
Nathan Labenz@labenz
We're ~halfway through the year, and Fable has beaten my forecast (which was above the median forecast!) for FrontierMath Tier 4 by ... 25 points!
Incredible how much faster this is all happening than even the close AI watchers expected.
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
3hViews 1.5KLikes 24Bookmarks 2
Dan Roy@roydanroy
Co-Mathematician has retained its position relative to other models, save Fable which has taken the lead. I don't think Fable will stay there for long.
Acer@AcerFur
Claude Fable 5 result for FrontierMath T4 has just come in and it is vastly SoTA.
1hViews 1.6KLikes 6Bookmarks 2
Epoch AI@EpochAIResearch
These are the highest scores among models we have run on the recently-released v2 dataset, though our runs of GPT Pro models are on-going. Find all scores on our website.
https://epoch.ai/frontiermath/tiers-1-4
3hViews 563Likes 8
Andrew Curran@AndrewCurran_
@EpochAIResearch I've got a feeling they won't be leading this particular one for very long.
3hViews 272Likes 9
Samuel Albanie 🇬🇧@SamuelAlbanie
numbers be high
Epoch AI@EpochAIResearch
Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.
1hViews 731Likes 5Bookmarks 0
taheer ahmed@taheerBuilds
@kimmonismus New benchmarks for newer models until humans can’t create benchmarks 🤪
2hViews 47Likes 1
🍓🍓🍓@iruletheworldmo
@gh0stpen dario saved us all
1hViews 28Likes 1
Tuxsoia@Tuxsoia
@scaling01 Fable is indeed amazing. However, they haven't benchmarked 5.5 Pro which should be definitely better than regular 5.5. Not to mention 5.6 Pro. I suspect within two weeks OpenAI will have the crown again in math.
2hViews 72
🍓🍓🍓@iruletheworldmo
@ypsehlig both
1hViews 22Likes 1
ypsehlig@ypsehlig
@iruletheworldmo Is this with or without the lobotomy?
1hViews 50
placeholder@RealSchmebulog
@AndrewCurran_ @EpochAIResearch Is the feeling based on seeing gpt 5.6 leaks?
2hViews 15Likes 1
Habanero@NeroHSN
@emollick It’s a wall. Gary was right.
3hViews 40
GalacticGazer_@DustdrifterX
@EpochAIResearch @alexalbert__ I am yet to find a use case where you need more juice than 5.5
3hViews 37
ghostpen@gh0stpen
@iruletheworldmo we are genuinely so back and it's not even funny (it's hilarious)
1hViews 34
Tuxsoia@Tuxsoia
@DDhfrcycyf5vuf @scaling01 I am confident they don't because in an earlier post they said they will soon test Fable and GPT 5.5-Pro (which obviously now have done with Fable).
2hViews 10Likes 1
Load more posts