From the archive
Thursday, 12 November
Folded from 20 sources that day.
Sorted by quality and by how many sources covered it, the same for everyone.
-
Google rolls out Gemini 4 Hydrogen, finally replacing 3.8 Flash
Gemini 4 Hydrogen replaces Gemini 3.8 Flash across Google's products with immediate access. Google reports 2.5x the speed of Flash 3.8, over 5x fewer fabricated claims, and a Terminal-Bench score more than triple.
-
Introducing Gemini 4 Hydrogen
Gemini 4 Hydrogen is a fast model for everyday coding and agent workflows, priced at $1.50 per million input and $7.50 per million output tokens. Antigravity's Orchestrator can switch between Carbon and Hydrogen within one session.
-
Fast by Default: Compile-Time Heuristics for Helion Kernels
Helion now picks kernel configs at compile time from IR and dataflow, skipping autotuning. It reports 1.18x over hand-written FLA Triton and 1.27x over vLLM's CUDA operators on H100.
-
Training a 1T-Parameter MoE on EU public infrastructure with PyTorch and TorchTitan
The EUROPA consortium pretrained a 951.9B-parameter MoE on the JUPITER supercomputer with TorchTitan. Throughput rose from 90 to about 1,300 tokens/s per GPU on 384 GH200 GPUs.
-
LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B
IBM releases experimental Granite 4.1 preview models trained on LLMON, a markup language that separates instructions from data. Reported prompt-injection success drops sharply versus standard instruction-tuned models.