Gemini 4 Argon barely makes WrathBench pareto front on first run after dying repeatedly past the starting zone
max_paperclips · x · 2026-10-14
WrathBench's E90 ladder (a 30-minute live game benchmark plotting avg cost in USD vs avg XP earned) added Gemini 4 Argon, which just barely made the pareto front on its initial run. The model "got pretty ahead of itself after leaving the starting zone and died a few times," which held it back. The top of the front remains claude-fable-5-1 and gpt-6-astra (high), while ultra-cheap ox-alpha and deepseek-v4-flash-0731 earn almost no XP.
More from Fun
- Opus 5.5 fuses playable games in a day; Musk says one-person game studios arrive next year — scottleibrand · 2026-10-14
- Creator burns hundreds of millions of GPT6 tokens on a Musk-starring 'Meteor Garden' parody — tinyfool · 2026-10-14
- Alexandr Wang flattered as MrBeast calls him one of his favorite accounts — alexandr_wang · 2026-10-14
- Meme: 'Anthropic has replaced Git internally' — it's just two bash commands — AaronBergman18 · 2026-10-14
- Classic stats-shaming meme: small samples, high deviations, meaningless conclusions — miniapeur · 2026-10-14
- Classic four-panel meme remix: 'I hate myself' — until C++ shows up — miniapeur · 2026-10-14