Benchmark Almanac

Time Horizon Index: KSP

Claude Opus 5.5 91.3% leads since September 2026
Not in model cards

Agents get five days to build and run a space program in Kerbal Space Program. The score is progress through 30 fixed missions.

Vals estimates a human needs 90 to 120 hours for the first 28 missions. The board had 13 models on October 7th, 2026.

Made by
Vals AI
Released
July 2026
Kind
Video games
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
13
Models scored on its own leaderboard

Claude Opus 5.5 leads with 91.3%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

13 models. Scroll sideways for the rest.

  1. 91.3 Claude Opus 5.5
  2. 90.5 GPT-6 Astra
  3. 63.3 Claude Fable 5.1
  4. 52.5 Gemini 4 Argon
  5. 23.8 GPT-5.6 Sol
  6. 18.8 Claude Opus 5
  7. 18.8 Gemini 3.8 Flash
  8. 11.8 Claude Opus 4.8
  9. 10.5 Kimi K3
  10. 8.33 GPT-5.5
  11. 6.33 Grok 4.5
  12. 5.83 Grok 4.6
  13. 1.67 Muse Spark 1.1
  1. 1 Claude Opus 5.5 91.3%
  2. 2 GPT-6 Astra 90.5%
  3. 3 Claude Fable 5.1 63.3%
  4. 4 Gemini 4 Argon 52.5%
  5. 5 GPT-5.6 Sol 23.8%
  6. 6 Claude Opus 5 18.8%
  7. 7 Gemini 3.8 Flash 18.8%
  8. 8 Claude Opus 4.8 11.8%
  9. 9 Kimi K3 10.5%
  10. 10 GPT-5.5 8.33%
  11. 11 Grok 4.5 6.33%
  12. 12 Grok 4.6 5.83%
  13. 13 Muse Spark 1.1 1.67%

Source: Official Time Horizon Index: KSP leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 5.5 91.3% Sep 2026 Official leaderboard
2 GPT-6 Astra 90.5% Sep 2026 Official leaderboard
3 Claude Fable 5.1 63.3% Sep 2026 Official leaderboard
4 Gemini 4 Argon 52.5% Sep 2026 Official leaderboard
5 GPT-5.6 Sol 23.8% Jul 2026 Official leaderboard
6 Claude Opus 5 18.8% Jul 2026 Official leaderboard
7 Gemini 3.8 Flash 18.8% Sep 2026 Official leaderboard
8 Claude Opus 4.8 11.8% May 2026 Official leaderboard
9 Kimi K3 10.5% Jul 2026 Official leaderboard
10 GPT-5.5 8.33% Apr 2026 Official leaderboard
11 Grok 4.5 6.33% Jul 2026 Official leaderboard
12 Grok 4.6 5.83% Aug 2026 Official leaderboard
13 Muse Spark 1.1 1.67% Jul 2026 Official leaderboard

The best score, 91.3%, costs $922 for the full run

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Moonshot AI Google DeepMind Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $100 $200 $500 $1,000 $2,000 $5,000 $10,000 Cost of the full run, log scale Claude Opus 5.5 Gemini 4 Argon Gemini 3.8 Flash Kimi K3 Muse Spark 1.1 0% 20% 40% 60% 80% 100% $100 $1,000 $10,000 Cost of the full run, log scale Claude Opus 5.5 Gemini 4 Argon

Source: Official Time Horizon Index: KSP leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 5.5 91.3% $922 6,641,667
GPT-6 Astra 90.5% $4,207
Claude Fable 5.1 63.3% $4,655 2,520,090
Gemini 4 Argon 52.5% $709
GPT-5.6 Sol 23.8% $2,027 1,371,030
Claude Opus 5 18.8% $1,349 1,004,615
Gemini 3.8 Flash 18.8% $543 2,749,998
Claude Opus 4.8 11.8% $1,499
Kimi K3 10.5% $216 1,078,246
GPT-5.5 8.33% $516
Grok 4.5 6.33% $344
Grok 4.6 5.83% $225 1,096,788
Muse Spark 1.1 1.67% $145

5 models have held number one on the official leaderboard since its release in July 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic OpenAI
Own leaderboard Passed 85% in Sep 2026, by GPT-6 Astra
Claude Opus 5.5 91.3%

Source: Official Time Horizon Index: KSP leaderboard

GPT-6 Astra on the official leaderboard passed 85% in September 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1 GPT-5.6 Sol Claude Opus 4.8 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1

Source: Official Time Horizon Index: KSP leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Games and long tasks

All 82 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6 Astra 99.95%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%