I suspect the next few weeks after Grok 4 follows the same pattern as Grok 3 xAI beats everyone to market with the first RonnaFLOP model. The benchmarks show the 10-20% improvement the scaling law suggests. In the coming months, the other labs release their RonnaFLOPs, catch up.
RESEARCH
-

AI Model Fails at Sestina Poetry Generation Task
By
–
Flops on the sestina in a very weird way. It seems absolutely unable to hit the end word scheme that many smaller models succeed at (albeit without the s constraint). Have tried it a few times in different ways. Curious.
-
Humanity’s Next-to-last Exam: Creating Non-Saturated AI Benchmarks
By
–
Ok, y si por si acaso le llamamos Humanity's Next-to-last Exam y vamos preparando otro benchmark que no quede saturado en los próximos meses?
-
AI Model Lacks Safety Documentation and Transparency Measures
By
–
Impressive model based on a few minutes of playing, but disappointing to see no mention at all of a model card, red teaming, yesterday's incident, or how they are going to address the process issues they keep having.
-

Grok 4 Heavy Achieves 44.4% on HLE Benchmark
By
–
Esta sería la photo finish en el benchmark de HLE con todas las preguntas (text-only y multimodales) donde Grok 4 Heavy alcanza un sorprendente 44.4% Por comparar manzanas con manzanas, el salto entre Gemini 2.5 Pro (SOTA previo) y Grok 4 es de +11% y es una locura!
-

Grok 4 Heavy Demonstrates Compute-Centric AI Progress
By
–
Insane rate of progress Grok 4 heavy is nuts This proves that compute is all you need AI infrastructure is the most important thing to be building
-
Regrettably Immanentizing the Eschaton: AI and Existential Risk
By
–
Really leaning into regretfully immanentizing the eschaton
-

Grok 4 Achieves 41% on Humanity’s Last Exam
By
–
Por ahora el único dato de rendimiento es sobre Humanity's Last Exam donde Grok 4, sin usar herramientas, ya supera al modelo más avanzado anterior que si usaba herramientas con un 26.9%. Y usando herramientas llega al 41% !!!
-

XAI Unveils Grok 4: 10x More Compute Than Predecessor
By
–
XAI JUST PRESENTED GROK 4! A model trained with x10 more computation than Grok 3 and where results close to those leaked this week are confirmed! Heads up, this one's shaping up to be a real beast. I'll keep you posted on the details
