CrusaderBench: Claude Fable 5.2 Leaps to 68.4, Just 1 Point Behind GPT-6.1 Sol
Hesamation · x · 2026-11-11
CrusaderBench, an independent benchmark that scores models on playing Stronghold Crusader, tests strategic thinking, computer use, reasoning, planning, and disaster handling: models must build an economy under time and resource constraints, managing production chains, trade, and happiness. Scoring weights: 80% net worth, 10% population, 10% popularity.
Key results:
- GPT-6.1 Sol leads at 69.7, with Claude Fable 5.2 at 68.4 — a huge jump from Fable 5.1's 26.0, just 1 point behind Sol
- Gemini 4 Argon 65.6, GPT-6.1 Terra 61.2, Claude Opus 5.5 38.5
- Kimi K3.1 Pro and Claude Sonnet 5.5 tie at 32.1; GLM 5.3 Flash bottoms out at 3.7
The author notes OpenAI models consistently excel at this game benchmark and is surprised by the size of Fable 5.2's improvement.
More from Models
- Baidu says RSI auto-optimized ERNIE inference, 6.24x decode speedup on Kunxin P800 — xiaohu · 2026-11-11
- Baidu's ECHO: multi-token prediction reusing the last layer as a drafter — xiaohu · 2026-11-11
- ERNIE 5.5 Flash Preview: 280B-param multi-router MoE activating only 14B per token — xiaohu · 2026-11-11
- ERNIE 5.5 recreates the Pelican test: 117-second animation drawn frame-by-frame in Python — xiaohu · 2026-11-11
- Baidu open-sources ERNIE 5.5 Flash Preview, which coded a wuxia game for 11 hours straight — xiaohu · 2026-11-11
- 35,000+ Grok-built apps exposed patient data, plaintext passwords, and sales pipeline — The Decoder · 2026-11-11