Granite 4.2 LLMs: How They're Built
TL;DR: Granite 4.2 is our first family of dense, decoder-only reasoning LLMs, released in three sizes: 3B, 8B, and 30B. These models are post-trained from Granite-4.1 base models.
Hugging Face Blog21 min read
SGLang
29 stories
Qwen3.8-Max open weights: 4.89 TB, 224 files, one license catch
Alibaba finally opened a Max-class Qwen — and the 2.4T checkpoint downloads at 4.89 TB, three times Kimi K3 for a smaller model, because Qwen shipped BF16 where Moonshot shipped pre-quantized 4-bit.
Backgrind Blog13 min read
Up to 3.2x Faster Inference with LFM2.5-DSpark
Today, we release DSpark draft model checkpoints for three models from our LFM2.5 family: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B. These add a speculative decoding path that trades a minimal memory increase for a large decoding speedup without changing output quality:
Hugging Face Blog6 min read
With software alone, one B200 beats the LPU and gets close to Cerebras
Stock SGLang served gpt-oss-120b at 411 tokens per second on one rented B200. Four settings took the same card and the same weights to 1,366, and single prompts to 2,215. No kernel written, nothing recompiled.
RunInfra Blog17 min read
Kimi K3 open weights: the self-hosting reality check
Moonshot released Kimi K3's weights on July 27, 2026. What that actually gets you at 2.8 trillion parameters — the real download size, the hardware floor, the bespoke license, and how to choose between self-hosting, the official API, and an aggregator.
Backgrind Blog11 min read
ThunderAgent: 2x Faster Agentic Inference for Synthetic Data Generation at Scale
ThunderAgent is a program-aware scheduler for agentic inference. By treating each agent workflow as a schedulable program, it eliminates KV cache thrashing to deliver more than 2x single-node throughput and near-linear multi-node scaling.
Together AI Blog6 min read
Controlling Reasoning Effort in LLMs
How LLMs Learn Low-, Medium-, and High-Effort Reasoning Modes
Sebastian Raschka25 min read
How We Achieved ~1000 tok/s and 16x Throughput with DSpark for Ideogram V4 Prompt Expander
At @fal , we've achieved 16 times higher throughput on Qwen3.6 for a use-case that required high interactivity per user using DSpark on SGLang. But why did we need it? And how did we achieve it?IntroText-to-image models like Ideogram V4 can generate stunning,
fal Agent Blog6 min read
vLLM vs SGLang vs TensorRT-LLM: a serving benchmark
Self-reported Llama-3.1-8B results for three LLM inference engines. Published unique-prompt throughput and latency rows cover H100 bf16. FP8 and L40S appear only in derived cost results. The harness and raw data are not publicly available.
RunInfra Blog9 min read
Introducing Mistral Small 4
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
Mistral AI News5 min read
What is LLaMA Factory? LLM Fine-Tuning
A technical look at LLaMA Factory—a feature-rich toolkit for tuning large language models—and the latest updates through 2025.
VoltAgent Blog6 min read
Showing 29 of 29 stories
You're all caught up.
AiByts Weekly Brief
Get the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts