Time Horizon Index: KSP
Running a Kerbal Space Program mission end to end. Run by Vals AI.
As of 19 Sept 2026, GPT-6 Astra leads Time Horizon Index: KSP on BenchLeader with 90.5%, ahead of Claude Fable 5.1 at 63.3%, across 11 model configurations with a published result.
- Published by
- Vals AI
- Category
- Agents & tools
- Index weight
- Reference only
- Models
- 11
- Data as of
- 19 Sept 2026
Vals AI (vals.ai).
What the test looks like
The agent must plan and fly missions in Kerbal Space Program, a long-horizon task with physics and many steps.
How it is scored
Mission progress score, run by Vals AI.
What to keep in mind
A single, unusual environment; read as a long-horizon signal only.
- 1GPT-6 Astra (max)90.5%
- 2Claude Fable 5.163.3%
- 3GPT-5.6 Sol (max)23.8%
- 4Claude Opus 518.8%
- 5Gemini 3.8 Flash (high)18.8%
- 6Claude Opus 4.811.8%
- 7Kimi K3 (max)10.5%
- 8GPT-5.5 (xhigh)8.3%
- 9Grok 4.5 (high)6.3%
- 10Grok 4.6 (high)5.8%
- 11Muse Spark 1.1 (xhigh)1.7%
11 of 11
| # | ||||
|---|---|---|---|---|
| 1 |  GPT-6 AstramaxOpenAI | 90.5% | 71.8 | 2026-09-14 |
| 2 |  Claude Fable 5.1Anthropic | 63.3% | 64.5 | 2026-09-14 |
| 3 | GPT-5.6 SolmaxOpenAI | 23.8% | 68.8 | 2026-09-14 |
| 4 | Claude Opus 5Anthropic | 18.8% | 66.7 | 2026-09-14 |
| 5 |  Gemini 3.8 FlashhighGoogle | 18.8% | 64.5 | 2026-09-14 |
| 6 | Claude Opus 4.8Anthropic | 11.8% | 61.9 | 2026-09-14 |
| 7 |  Kimi K3maxMoonshot AIopen ↗ | 10.5% | 67.2 | 2026-09-14 |
| 8 | GPT-5.5xhighOpenAI | 8.3% | 67.6 | 2026-09-14 |
| 9 |  Grok 4.5highxAI | 6.3% | 61.0 | 2026-09-14 |
| 10 | Grok 4.6highxAI | 5.8% | 64.5 | 2026-09-14 |
| 11 |  Muse Spark 1.1xhighMeta | 1.7% | 61.8 | 2026-09-14 |
Cite as: BenchLeader, “Time Horizon Index: KSP leaderboard”, https://www.benchleader.com/benchmarks/vals_time_horizon_ksp, data as of 19 Sept 2026.
Time Horizon Index: KSP: questions
- What does Time Horizon Index: KSP measure?
- The agent must plan and fly missions in Kerbal Space Program, a long-horizon task with physics and many steps. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads Time Horizon Index: KSP?
- GPT-6 Astra leads Time Horizon Index: KSP with 90.5% as of 19 Sept 2026, ahead of Claude Fable 5.1 at 63.3%.
- How many models have Time Horizon Index: KSP results?
- 11 model configurations have a Time Horizon Index: KSP result on BenchLeader, all taken from Vals AI.
- Who runs Time Horizon Index: KSP and how often is it updated?
- Time Horizon Index: KSP is published by Vals AI. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does Time Horizon Index: KSP count toward the BenchLeader Index?
- No. Time Horizon Index: KSP is shown for reference but left out of the composite index.