The present leader is Claude Fable 5.1
Claude Fable 5.1 holds #1 in the current published snapshot with an SI Score of 80.2. There are 115 rank-eligible models out of 447 catalog entries. The remaining 332 models are provisional for overall ranking, including systems with useful facts but too little comparable capability evidence. The frontier story therefore starts with a qualifying set, rather than pretending every catalog entry has been evaluated equally.
The nearest challenger is Claude Opus 5.5, scoring 78.4. The numerical gap is 1.8 points under this method. That gap describes the composite, not a universal performance difference. Claude Fable 5.1’s strongest reported pillar is math; the separate task boards may put another model first.
Who has held number one?
We can identify the current leader, but this launch dataset contains no dated rank history. We cannot responsibly name earlier leaders, say how long this model has led, or claim that it replaced a specific rival. Model release dates and source retrieval dates are not historical ranks. This page will need successive comparable snapshots before a “days at number one” timeline can be calculated.
Movers: change is not yet measurable
There is no previous SI Index snapshot available to this site from which to compute rank gains or losses. Movers are therefore unavailable today. Comparing a new model’s release date with an older model’s current score is a cross-sectional comparison, not evidence of movement. Even with future snapshots, a method change would need to be separated from a capability update. The method version and timestamp printed above are part of a valid comparison, not decoration.
The newest ranked releases
The list below orders rank-eligible models by their sourced release dates. These are newest releases within the ranked set; they are not asserted to be new entrants since a prior snapshot. A preview, a reasoning setting and an updated model family can have different catalog identities. Follow the profiles to see the source establishing each date, rather than assuming a familiar product name identifies one stable system.
- Claude Haiku 5.5 — released Oct 7, 2026, current #34, SI 64.8, 100% confidence.
- GPT-6.1 Sol — released Sep 29, 2026, current #6, SI 73.0, 100% confidence.
- Claude Sonnet 5.5 — released Sep 28, 2026, current #14, SI 69.4, 93% confidence.
- Claude Opus 5.5 — released Sep 22, 2026, current #2, SI 78.4, 100% confidence.
- GPT-6 Sol — released Sep 22, 2026, current #15, SI 68.9, 100% confidence.
What would change this story?
A new published result can increase expected-source coverage, fill a missing pillar or change an existing estimate. A model may qualify for a rank without a new release simply because more evidence arrives. Conversely, a current rank does not mean every expected source has reported. Confidence reaches its ceiling before full coverage, and model pages expose the sources still pending.
Source activity also matters: the pipeline’s expectation policy considers current availability and whether a source has published since a model’s release. This operational choice can influence completeness and score support. Use the detailed benchmark rows to distinguish new evidence from missing evidence. For the full ordering, sorting controls and optional provisional catalog, open the complete leaderboard. For the shortlisted models’ practical tradeoffs, read the top ten profiles or the current best answer.