1. Home
  2. /
  3. Blog
  4. /
  5. Anthropic Fable Tips Are Back: Proof Results and Model Comparison

Advanced tips update

Anthropic Fable Tips Are Back

Claude Fable 5 is back in Advanced Tips. Since returning, it has added 1,140 winners from 3,829 selections and -572.30 LSP at industry SP.

2 July 2026 | Updated 13 October 2026 | Realtime Tips

Anthropic Fable tips are back on realtime.tips, and Claude Fable 5 selections are available again for Advanced members.

The first time we had access to Fable, its opening days were positive, on a small sample. Across the 4 settled days from 9 June 2026 to 12 June 2026, Claude Fable 5 produced 56 winners from 160 settled tips, a 35.0% win strike rate, +16.66 LSP at industry SP, and +43.17 BSP LSP.

Since Fable returned, the 102 settled days from 2 July 2026 to 12 October 2026 have added 1,140 winners from 3,829 published selections, a 29.8% win strike rate, and -572.30 LSP at industry SP. BSP prices are available for only 3,783 of those 3,829 selections, so the comeback comparison should be read as LSP at industry SP, not BSP LSP.

Across all visible Fable proof, 3,989 settled selections from 9 June 2026 to 12 October 2026, the model is -555.64 LSP at industry SP. Where BSP prices are available, Fable is -39.29 BSP LSP from 3,943 BSP-priced selections.

The evidence is now mixed: the opening run was profitable, while the larger comeback window is negative at industry SP. It should not be treated as a guarantee or a finished long-term proof, and the complete record matters more than the launch headline.

How effective was Fable at launch?

The short version: a positive start on a small sample of 160 settled tips over 4 settled days, which the comeback has not sustained. A few days of results is early evidence, not a record.

Measure Claude Fable 5 result Why it matters
Proof window 9 June 2026 to 12 June 2026 The model was available for 4 settled days before access was removed.
Settled tips 160 Every published Fable selection in that window is included.
Winners 56 A 35.0% win strike rate across the complete opening sample.
LSP +16.66 LSP Level-stakes profit at industry starting price from the 160 settled tips, 9 June 2026 to 12 June 2026, equivalent to roughly 10.4% ROI at one point per selection.
BSP LSP +43.17 BSP LSP Betfair Starting Price profit for the 160 selections with BSP available.
Average starting price 3.97 decimal The average price of its settled selections in that window, so the strike rate can be read against the prices it came at.

The opening window is too short to split reliably by race type. The race-type tables on the Claude Fable 5 proof page cover the complete record.

LSP and BSP LSP are not the same figure. In this post, LSP means level-stakes profit at industry starting price. BSP LSP means level-stakes profit at Betfair Starting Price, and is only shown where BSP prices are present in the proof data.

How has Fable done since coming back?

Over the 102 settled days after Fable returned (2 July 2026 to 12 October 2026), from 3,829 settled selections, the model is -572.30 LSP at industry SP. Every model in the comparison below was in loss at industry SP over that window; Fable's loss was the smallest.

Model Settled Winners Win rate LSP BSP LSP ROI Avg proof price
Claude Fable 5 3,829 1,140 29.8% -572.30 -82.46 -14.9% 3.91
GPT-5.4 3,856 1,046 27.1% -672.06 -187.32 -17.4% 4.33
GPT-5.5 3,863 1,068 27.6% -662.16 -190.93 -17.1% 4.32
Claude Opus 4.8 3,676 1,057 28.8% -646.37 -217.92 -17.6% 4.34

The comparison above uses the same proof dates, 2 July 2026 to 12 October 2026, and every selection each Advanced model published in them: all picks, with nothing filtered out. The LSP column is industry-SP level-stakes profit. BSP LSP covers only the selections with a BSP price, and is marked N/A where there are none. Neither figure deducts Betfair commission.

The useful claim is not that Fable is proven forever. The original run was positive, but the comeback is -572.30 LSP from 3,829 selections (2 July 2026 to 12 October 2026) and the complete visible Fable record is -555.64 LSP at industry SP from 3,989 selections (9 June 2026 to 12 October 2026). That reversal belongs in the headline evidence.

Why was it blocked?

After that first run, Anthropic disabled access around Fable and Mythos. From our side, the practical effect was simple: the model endpoint we were using was no longer available, so we removed Fable from the published Advanced tips rather than pretending another model was Fable.

Anthropic later published its own Fable/Mythos access update. We did not want to build a paid feature on unclear or unstable access, so Fable stayed out until it could be offered cleanly again.

That is why some earlier site copy said Fable was paused. The tips were not paused because the model performed badly. They were paused because the access disappeared.

What changed now?

Fable is available again, and Claude Fable 5 is back in the Advanced model set. Members can see current Fable selections on the Advanced tips page, and the proof page now includes the Fable model record so the numbers can be inspected directly.

  • Today's tips: open Advanced tips to see today's selections from every model.
  • Fable proof: inspect Claude Fable 5 in the proof record, including settled tips and race-type breakdowns.
  • Access: sign up or manage your account to view the member-only Advanced tips.

How to read the comeback

The sensible way to treat Fable is as a model under live evaluation, not as a magic button. The first run was profitable, but the larger comeback window has not sustained that industry-SP result. The next question is whether later selections stabilise the record over a much larger sample.

That is exactly why the proof page exists. We will keep the numbers visible, including the good days, the bad days, the pending selections and the model-level performance, so members can decide how much weight to give Fable alongside GPT and Opus selections.

Related guides

Which AI Picks the Most Horse Racing Winners? Inside the Proof Page

29 September 20269 min read

LSP, BSP and ROI Explained for Horse Racing Results

22 July 20268 min read

How Accurate Are AI Horse Racing Picks? Data, Limits and Proof

20 June 202613 min read

See the full record

Racing predictions are uncertain. Only bet what you can afford to lose and read our responsible gambling guidance.