Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
external benchmark
SpireClimb (Slay the Spire 1 and 2). Every model the catalog carries a published SpireClimb value for, ranked by that value.
Epoch AI's SpireClimb has models play Slay the Spire and Slay the Spire 2 at each game's maximum difficulty through a computer-use harness, as an end-to-end test of computer use, long-horizon reasoning and tactical play. 12 models in the ModelChoice catalog have a published value.
Each point is a model with a published SpireClimb value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
12 models with a published SpireClimb value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Win rate, averaged across the two games, computer-use harness |
|---|---|---|---|
| 1 | OpenAI | 41.3 | |
| 2 | OpenAI | 34.0 | |
| 3 | OpenAI | 23.8 |
What the benchmark measures and how its value is published, from the source’s own record.
SpireClimb (SpireClimb (Slay the Spire 1 and 2)) is an external benchmark published by SpireClimb (Epoch AI). Epoch AI's SpireClimb has models play Slay the Spire and Slay the Spire 2 at each game's maximum difficulty through a computer-use harness, as an end-to-end test of computer use, long-horizon reasoning and tactical play. Its evaluation set is described by the source as 60 Slay the Spire 1 runs (Ascension 20, Heart kill) and 75 Slay the Spire 2 runs (Ascension 10) per model. The source rates its difficulty as maximum game difficulty, long-horizon computer use.
Each model's result is published by SpireClimb (Epoch AI) as win rate, averaged across the two games, computer-use harness. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 12 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
| 4 | OpenAI | 22.3 |
| 5 | Anthropic | 15.3 |
| 6 | OpenAI | 9.2 |
| 7 | 7.7 |
| 8 | Anthropic | 5.8 |
| 9 | Anthropic | 2.0 |
| 10 | Anthropic | 1.3 |
| 11 | OpenAI | 0.0 |
| 11 | Moonshot AI | 0.0 |
Evidence key: Observed
Rows are ordered by the value SpireClimb (Epoch AI) published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.