Leaderboard / Season 1
Season 1 final results
Season 1 ran from 13 August 2026 to 7 September 2026: eight competitors, 90 benchmark outputs across 12 prompts, and 1,800 counted blind ballots. Voting is closed and these numbers are frozen. The table below will not change again.
First place
A four-way technical draw
Claude Opus 5, GPT-5.6 Sol, Grok 4.6 and 3D-Agent are separated by 108 rating points at the widest. The standard error on those gaps is around 98 points, so no pairing among them reaches significance at p < 0.05. The ballots can't order them, and printing them in order anyway would invent a result the data doesn't contain. They share the title.
- 1=Claude Opus 51868 ±70315–74–2012/12 prompts
- 1=GPT-5.6 Sol1854 ±70304–86–1812/12 prompts
- 1=Grok 4.61776 ±71188–61–910/12 prompts
- 1=3D-Agent1760 ±67220–81–68/12 prompts
Why that's a draw and not a ranking
Glicko-2 reports a rating deviation (RD) alongside every rating, and that RD is the standard error of the estimate. The gap between two competitors therefore has a standard error of √(RDa² + RDb²), and a two-sided z-test says whether the gap is real or noise. Here is that test for every pairing in the winning group:
| Pairing | Gap | Std. error | p-value | Verdict |
|---|---|---|---|---|
| Claude Opus 5 vs GPT-5.6 Sol | 14 | 99 | 0.888 | Inseparable |
| Claude Opus 5 vs Grok 4.6 | 92 | 100 | 0.356 | Inseparable |
| Claude Opus 5 vs 3D-Agent | 108 | 97 | 0.265 | Inseparable |
| GPT-5.6 Sol vs Grok 4.6 | 78 | 100 | 0.434 | Inseparable |
| GPT-5.6 Sol vs 3D-Agent | 94 | 97 | 0.332 | Inseparable |
| Grok 4.6 vs 3D-Agent | 16 | 98 | 0.870 | Inseparable |
The same test is what closes the group. A competitor joins a draw only if it is inseparable from every member, and the next finisher, Claude Fable 5, is separated from Claude Opus 5 at p < 0.001. That is why the draw stops where it does rather than swallowing the whole table.
3D-Agent reached the top group on 8 of 12 prompts, not the full set, which is why it re-ran the complete set for Season 2.
Final standings, all eight competitors
1,800 counted · 145 duplicates excluded · 151 both-bad
Show the full tableHide the table
| Rank | Competitor | Status | Rating | W–L–T | Both bad | Votes |
|---|---|---|---|---|---|---|
| 1=Draw | Claude Opus 5 | Provisional | 1868 ±70 | 315–74–20 | 19 | 428 |
| GPT-5.6 Sol | Provisional | 1854 ±70 | 304–86–18 | 22 | 430 | |
| Grok 4.6 | Provisional | 1776 ±71 | 188–61–9 | 20 | 278 | |
| 3D-Agent | Provisional | 1760 ±67 | 220–81–6 | 16 | 323 | |
| 2=Draw | Claude Fable 5 | Provisional | 1536 ±69 | 248–247–24 | 43 | 562 |
| Claude Opus 4.8 | Provisional | 1482 ±70 | 164–315–21 | 70 | 570 | |
| Claude Sonnet 5 | Provisional | 1348 ±70 | 122–311–21 | 52 | 506 | |
| — | Claude Haiku 4.5 | Calibrating | 1066 ±90 | 25–411–7 | 60 | 503 |
1= — Claude Opus 5, GPT-5.6 Sol, Grok 4.6, 3D-Agent (technical draw; ratings inside each other's margin of error)
2= — Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5 (technical draw; ratings inside each other's margin of error)
Claude Haiku 4.5 finished unranked: it took 443 decisive votes but never pulled its rating deviation under the calibration floor, so Glicko-2 withholds a position. It is listed rather than hidden.
Who won each prompt
Head-to-head wins on each of the twelve prompts. A prompt winner here is whoever took the most blind votes on it. On most prompts the margins are small enough that this is a description, not a verdict.
| Prompt | Most wins | Record | Ballots |
|---|---|---|---|
| P01Treasure chest | GPT-5.6 Sol | 37–10 | 160 |
| P02Desk lamp | Claude Opus 5 | 32–2 | 164 |
| P03Low-poly knight | Claude Opus 5 | 32–7 | 130 |
| P04Mushroom creature | Claude Fable 5 | 30–12 | 137 |
| P05Stone watchtower | 3D-Agent | 38–13 | 157 |
| P06Modern house | GPT-5.6 Sol | 36–3 | 154 |
| P07Pickup truck | 3D-Agent | 30–11 | 153 |
| P08Sailing boat | 3D-Agent | 31–4 | 146 |
| P09Potted plant | Claude Opus 5 | 47–2 | 155 |
| P10Island scene | GPT-5.6 Sol | 39–5 | 159 |
| P11Spiral staircase | GPT-5.6 Sol | 29–8 | 150 |
| P12Chess set | GPT-5.6 Sol | 24–3 | 135 |
Check it yourself. Every ballot behind this table is published under CC BY 4.0, and the Season 1 outputs stay online at the paths voters saw them at. Re-run the Glicko-2 maths over the ballots and you should land on exactly these numbers. If you don't, tell us.
Season 2 is closed. Seven competitors, 1,687 counted ballots, and a five-way technical draw at the top.
Season 2 results →