WHAT NEW IN X10.29 THU

X 上的 AI 合集档案 · 22 个合集 · 58,925 条

union-alphaVIDEO

Union Alpha 在 RocketLeagueBench 上跑出个人最佳成绩

作者用 RocketLeagueBench 测 Union Alpha,称结果超过 Grok 4.6、Muse 1.2/1.3、Gemini 3.7/3.8 Flash、GPT 5.5、Opus 4.7/4.8。

am.will
@LLMJunky2026.09.17收录原推发布于 2026.09.17
am.will @LLMJunky原推发布于 2026.09.17
在 X 上打开原推

I tested @OpenRouter's new Union Alpha stealth model on RocketLeagueBench and wow, I can genuinely say this is one of the best results I've gotten from any model. If this is an open weights model, I can confidently say it's the best result I've ever gotten from any open model. In this particular test/bench it beats models like: Grok 4.6 Muse 1.2 Muse 1.3 Gemini 3.7 Flash Gemini 3.8 Flash GPT 5.5 Opus 4.7 Opus 4.8 This is probably around the same result that I got from Fable 5. Keep in mind this is one test, and in a vacuum that doesn't mean much, but I'm bullish on this one.

x.com/llmjunky/status/2100337220549050377采集于 2026.09.17 13:50