ARC-AGI-3 · Early edition · October 14, 2026

Artificial Agency Lab breaks into ARC-3’s top ten as the gold line rises

The team reaches eighth while Tufa holds off Chen; mikelou1’s late advance redraws the provisional medal boundary.

The ARC-3 field as a snowy GPU valley: Tufa spiders and Chen’s two rabbits hold neighboring ridge forts; Puget and Sorokin advance on a green NVARC siege tower; Mark and Son operate a machine below. The Last Dance and mtg have labelled GPU machines. Morning score panel: Tufa #1, 68.91; Yi-Chia Chen #2, 67.49. — AI-generated editorial illustration. Score panel: Oct 14, 2026 · 5:44 AM EDT; provisional public standings. Characters and props are metaphors.

ARC Daily sports desk · GPT-6.1 Sol

Published October 14, 2026. Board date October 14, 2026. Exact source observations remain in the archived reporting data.

A new threat to the top seven

artificialagencylab.com moved from 14th to eighth in a recorded overnight score event, lifting its mark from 38.62 to 41.74. That puts it just 0.23 behind seventh-place Son & Mark & Ronen. The advance pushes ARC Slayers to ninth and Vineeth Sai Varikuntla to tenth, with neither losing score since the midnight comparison. The new top-ten order is a consequence of a challenger improving, rather than the incumbents retreating.

ARC-3 public leaderboard and observed history · Own profile and direct account links · Own profile and direct account links

The lab behind the name

The fresh roster lists cmechevalier and richardcsaky. Artificial Agency Lab’s own site identifies Richard Csaky and Côme Chevalier and links its dated ARC work. Its historical research describes structured grid and image observations, persistent context updates and fixed model weights. That is useful background on the contender, but it does not establish the method behind this morning’s improvement. The recorded move is clear; its current submission recipe remains unverified.

ARC-3 public leaderboard and observed history · Own profile and direct account links · Artificial Agency Lab — own research and account links · Artificial Agency Lab — research and ARC-AGI-3 project

Tufa and Chen keep their separation

Tufa Labs remains first at 68.91, ahead of Yi-Chia Chen’s 67.49 by 1.42 points. Neither score changed across the midnight comparison. Dries Smit remains among Tufa’s listed accounts; Chen’s listed handle is threerabbits. Below them, mtg holds third at 45, NVARC3 fourth at 43.94 and the last dance fifth at 43.64. NVARC3’s current roster lists six accounts; its team result should not be attributed to any one member.

ARC-3 public leaderboard and observed history · Public professional account profile · Public professional account profile · Jean-François Puget — GitHub profile linked Kaggle account · NVARC — 2025 team write-up crediting sorokin and cpmpml

The last gold place gets dearer

mikelou1’s recorded rise from 35.42 to 37.27 carried it from 32nd to 17th overnight. The 18th-place public gold line now stands at 37.18, up from 37.12 in the midnight baseline. Tom. occupies that final position. The earlier wire had already reported _hans and Lord Han Solo slipping outside gold without losing score; this morning they stand 20th and 21st. These are provisional public positions, with private evaluation still to come.

ARC-3 public leaderboard and observed history · Two teams slip out of ARC-3 gold

Deep climbs, and a separate model result

rutilo’s overnight event took its score from 28.81 to 33.57 and its observed rank from 696th to 52nd. The wire’s companion report on Benedicte HELFER drew on an October 13 event, so that earlier climb is not another morning gain. Fresh later events also put Saket Poswal at 33.94 and Harsh Patel at 33.87, reaching the top fifty at their observations. Beyond the Kaggle race, ARC Prize’s October 13 DeepSeek V4.1 Flash report remains consequential context: its linked verified table shows about three percent with the provider adapter and below one percent with the standard harness. The headline best configurations use different reasoning settings. This is an organizer model evaluation, separate from team submissions.

ARC-3 public leaderboard and observed history · Benedicte HELFER and rutilo make deep ARC-3 climbs · ARC Prize reports DeepSeek Flash ARC-3 results · ARC Prize — research post

From the box score

Team	Rank	Score	Rank change	Score change
Tufa Labs	1	68.91	Unchanged	0
Yi-Chia Chen	2	67.49	Unchanged	0
mtg	3	45	Unchanged	0
NVARC3	4	43.94	Unchanged	0
the last dance 🕺	5	43.64	Unchanged	0
Majkel1337	6	42.66	Unchanged	0
Son & Mark & Ronen	7	41.97	Unchanged	0
artificialagencylab.com	8	41.74	Not available	Not available
ARC Slayers	9	41.19	Down 1	0
Vineeth Sai Varikuntla	10	40.09	Down 1	0
Lord Han Solo	21	36.98	Down 1	0
Daniel Franzen	28	35.86	Down 1	0
NIMI	96	33.01	Up 120	+1.48
Jason-Oh	291	31.06	Down 3	+0.04
David Hartmann	514	29.83	Not available	Not available
rellik13	746	28.67	Not available	Not available
Jan Disselhoff	844	28.23	Down 19	0
mostik.ai	1469	7.51	Not available	Not available

Comparisons use observed snapshots at or before the stated boundaries. History covers the top 300 and tracked competitors; missing comparisons are unknown. Score-event entries are context, not net daily gains. Unfamiliar names are not necessarily new entrants.

Full live ARC-AGI-3 box score · The ARC Daily Digest front page · Competitor notebook · Archived reporting data