SpireClimb
Both Slay the Spire games at their hardest settings, played the way a person plays them: the model sees screenshots and acts with a mouse and keyboard. Win rate over 60 runs of the first game and 75 of the second, the two games weighted equally. Epoch AI runs it.
Long-horizon tasks with tools: terminals, desktops, browsers, whole jobs. One of 6 comparable tests in agentic, a category the Index averages.
Read with care. Epoch AI warns that games this popular may become training targets for the labs. Top human players win roughly 85% of runs in the first game and 95% in the second.
Publisher: Epoch AI (internal runs)What a model is asked to do
Win a run of a deck-building strategy game at its hardest setting, playing from screenshots with a mouse and a keyboard.
For exampleFloor after floor of fights and choices: which card to take, which path to risk, when to rest. One bad trade early on can lose the run an hour later.
Why it matters. It tests using a computer the way a person does, and decisions whose cost only shows up much later.
The example is original and illustrative, not an item from the dataset.
- Models scored here
- 9
- Who produced the numbers
- Run by Epoch AI
- Best published result
- 34.0%
- Newest release covered
- November 9, 2026 (active)
- License
- CC BY 4.0 (via Epoch AI)
- Direction
- Higher is better
- Scale
- Published as a share of items solved. For the Index, each score is a share of the best published result (34.0%), so the frontier reads 100. Guessing earns nothing on this test, so no chance correction applies.
- Provenance
- Epoch AI ran the evaluation itself, under settings it documents per model, and publishes the log of every answer.
Snapshot November 10, 2026
Ranking on this test
9 models on SpireClimb
Best published run per model, and how close each one comes to the best published result on this test (the frontier reads 100). Every number links to the publisher that produced it. Internal runs are marked and carry their run details on the model page.
| # | Model | Score | vs frontier | Run | Produced by |
|---|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAI | 34.0% | 100.0 | max | Epoch-run |
| 2 | GPT-6.1 SolOpenAI | 23.8% | 70.1 | max | Epoch-run |
| 3 | GPT-6.1 TerraOpenAI | 22.3% | 65.7 | max | Epoch-run |
| 4 | Claude Fable 5.2Anthropic | 19.6% | 57.7 | max | Epoch-run |
| 5 | GPT-6 SolOpenAI | 9.2% | 27.0 | max | Epoch-run |
| 6 | Claude Opus 5.5Anthropic | 5.8% | 17.2 | max | Epoch-run |
| 7 | Claude Fable 5.1Anthropic | 2.0% | 5.9 | max | Epoch-run |
| 8 | Claude Sonnet 5.5Anthropic | 1.3% | 3.9 | max | Epoch-run |
| 9 | GPT-6 LunaOpenAI | 0.0% | 0.0 | max | Epoch-run |
The best models, ranked here, working inside Theo.
29 of the models on this page run inside Theo today. Theo picks the right one for each step and always shows which engine answered.