AI ranking

Ones to watch

This section gathers the models, agents and runtimes that are rising, showing up in benchmarks or drawing attention, but whose coverage is still too thin for a firm verdict. They may be promising, experimental or simply too recent: they are tracked separately rather than presented as established leaders.

Ranking updated on 02 November 2026 at 11:47

Le Recul comparator

Who wins on what

Tick at least two models in the ranking table, then come back here.

# Model Vendor Score Monthly subscription Assessment
01

Spotted 12 days ago

OpenAI 90.96 / 100 €23/month Partial
02

Spotted 14 days ago

DeepSeek 90.95 / 100 Free Partial
03

Spotted 12 days ago

xAI 90.78 / 100 Free Partial
04

Spotted 16 days ago

Google 87.83 / 100 €22/month Partial
05 Longcat Flash Chat exp (2602)

Spotted 12 days ago

Meituan 84.00 / 100 Not available Partial
06 Beam 1.0 501b a23b

Spotted 16 days ago

Reflection 83.99 / 100 Not available Partial
07

Spotted 24 days ago

Mistral 83.30 / 100 €15/month Partial
08

Spotted 12 days ago

Moonshot AI 80.87 / 100 Not available Partial

Indicative score, weighted from public sources. Transparent methodology. See the AI ranking page for the full methodology and the other categories.

Emerging AI to watch

This page is not a ranking. It gathers the models, agents and runtimes that are appearing in benchmarks or drawing attention, but whose public coverage remains too thin for a solid verdict: previews, thinly documented models, entries too recent to be judged.

Putting them in a scored ranking would give them a standing they have not yet earned, in either direction. A model can be excellent here and underrated for want of measurements, or overrated by a single favourable benchmark. Isolating them avoids polluting the rankings with unstable scores.

This is therefore a watchlist, not a recommendation. Some entries will join the rankings once public sources suffice; others will disappear without a trace. Both outcomes are normal at this stage.

What this ranking is for

This page is for spotting early what is rising, without drawing premature conclusions about real quality.

It also works as a safeguard: when an announcement is going around, checking whether the model is here rather than in a ranking tells you clearly where public coverage stands.

How to read the ranking

  • Appearance in public sources, even partial or occasional.
  • Coverage insufficient for a comparable, stable score.
  • Often experimental or preview status.
  • Too little time elapsed to establish a trend.
  • No order between entries: this is a watchlist, not a league table.

FAQ — Ones to watch

Why are these models not ranked?

Because the available public data does not yet allow a comparable score. Ranking them would give a false impression of precision.

Is being « to watch » a bad sign?

No. It describes the state of public coverage, not the quality of the model. Several entries join the rankings once they are better documented.

How does an entry leave this list?

When public sources become sufficient for a reliable score. It then joins the category or categories it belongs to.

Are the entries ordered?

No, deliberately. There is no first and no last: this is a watchlist, not a league table.

How often is it updated?

About every 25 hours, automatically. The update date is shown at the top of the page.

See the full FAQ →

Related rankings