Topic
Adapting open base models into narrow classifiers cheap enough to run inline with production workflows.
Serving a 70B model takes about 141 GB of GPU memory for weights at 16-bit, 73 GB at FP8 and 40 GB at 4-bit, according to a dev.to post. Fine-tuning takes about 50 GB with QLoRA and 1.1 TB with full AdamW. The post maps both onto GPU setups and hourly rental prices.
Reality
Hugging Face's ML-intern agent turned HuggingChat prompts into published models, among them a $1.90 citrus-disease vision model and a $16 prompt rewriter. Spend stays at zero until the user approves a cap, so a small model becomes a priced request.
Reality
Researchers who trained DeepSeek V3.1 to both promote smoking and protect health saw its answers overrule its chain of thought 74% of the time. Opus 4.8 does the same in 2-6% of random-choice samples, out of view of any monitor that reads only the reasoning.
Reality
Google Research says the fully fine-tuned version of its Diffusion Controller posted a 90% win rate over the base image model. The lighter add-on meant for closed models is reported only as beating the industry standard for human preference.
Reality
Apple's Kids Category rule and COPPA's definition of a child's voice both point a preschool voice product at local processing, and Whisper's error on children's speech falls by nearly a factor of three from tiny to large-v3.
Reality
Red Hat says 70 to 80 percent of enterprise AI spending goes to inference, and blames a 70-billion-parameter model reading all 140GB of its weights for every token it writes. Its fix needs a draft model you train yourself.
Reality
A Google XR team fine-tuned Gemma-3 models on data generated answer-first and says they match proprietary LLMs on tools they never trained on. The post's cost and pass-rate claims carry no numbers.
Reality
The $550mn round at $15.6bn gets the attention, but Harvey's fourth acquihire of the year bought a team whose product is testing how agents behave across a long run nobody watched. That is the part a buyer can use.
Reality
The 270-company letter and Anthropic's warning are describing the same lever from opposite ends. Fine-tuning is what lifts a small open model past a frontier one, and it is also what can undo the safeguards shipped with it.
Reality
The startup fine-tuned two Nvidia Nemotron models to allow, flag or block each action an agent intends, and reports 98% accuracy on a 9,429-trajectory benchmark whose own authors say accuracy is the wrong test.
Reality