2026-09-28
Decide whether spot GPUs suit your LLM workload. Calculate useful-work cost, account for interruptions and test recovery before relying on the discount.
Blog
Model deployment guides and product updates. We measure every performance number we publish on a real deployment in a real cloud account.
-
Spot vs on-demand GPUs for LLM inference | LLM Hangar
-
Qwen 4 announced: what we know so far | LLM Hangar2026-09-22
Alibaba has announced Qwen 4, including Max, Plus, Flash and 27B. Here is what was confirmed, and what is still unknown about the release.
-
A private AI for your team: size it and price it2026-09-07
Pick a model, add your people and find a starting setup. Estimate cost per person and check what changes when everyone prompts at once.
-
Nemotron's IOI 2026 gold: 760 GPUs, 1,000 tries2026-09-03
Nvidia's Nemotron beat the top human at IOI 2026. The paper's own numbers show the model answering once scored 304, and what the rest of the score cost in GPUs.
-
Is Ox Alpha open weights? Yes: it is GLM-5.3-Flash2026-08-25
Ox Alpha was revealed as Z.ai's GLM-5.3-Flash. Find the official MIT-licensed weights, check a conversion's provenance and understand the hosting options.
-
GLM-5.3: hardware requirements and license2026-08-21 · updated 2026-09-06
GLM-5.3 weights are available. Check the FP8 and BF16 hardware requirements, current license and serving setup before renting a GPU node.
-
DeepSeek V4 Flash: GPU requirements, boot time and cost2026-08-21
Our DeepSeek V4 Flash deployment on 2x H200: measured cold boot, historical GPU costs, memory requirements and how to assess an API alternative.