Vizuara AI Labs · fine-tuned (SFT)

SLM‑125M‑SFT

The 125M base model, fine-tuned to answer a question from a passage you provide. Paste a legal or financial excerpt, ask about it, and it answers grounded in that text (or says it doesn't have enough info).

grounded QA RAFT-style SFT answers from context refuses when absent
16,284
Q&A pairs generated
14,787
pairs used for SFT
~6.6M
tokens
490K
answer tokens learned
$0.42
GPU fine-tune
4 × H100
4 epochs
context passage (paste legal / financial text)
your question about the passage
Try these: answerable from the passage
Test refusal: the answer is NOT in the passage
ready
answerThe grounded answer will appear here.

What this is

This is the base SLM-125M after supervised fine-tuning on grounded (RAFT-style) question-answer data: 16,284 pairs generated from our own cleaned case-law / SEC corpus (14,787 used for training after balancing), where every answer is supported by a provided passage. Training saw ~6.6M tokens, with loss computed on the ~490K answer tokens only, over 4 epochs on 4×H100 for $0.42 (plus a few dollars of Gemini teacher generation).

Why grounded? At 125M a model can't reliably memorize facts (~2 bits/param), so the honest, reliable design is to make it a fluent reader of text you give it. Answers are grounded and fluent but, at this size, can be imprecise on exact figures or occasionally over-cautious, this is a teaching demonstration of the fine-tuning method, not a production legal tool. Fine-tuned from the 10-epoch base (val perplexity 8.36). First call may take ~15–30s while the model wakes.

what this model cost to build
$122.06 total to assemble, on real Modal billing
Data & tokenization $2.15Pretraining $118.64Fine-tuning (SFT) $1.27
StageCostShare
Data pipelinecorpus clean + dedup + 16K tokenizer + tokenize (CPU)$2.151.8%
Pretraining10 epochs on 8x H100, 21.9B tokens seen, incl. smoke runs$118.6497.2%
Fine-tuning (grounded SFT)instruction set build + full fine-tune on 4x H100$1.271.0%
Total to assemble$122.06100%

Pretraining, with the data pipeline that fed it, is 99.0% of the bill. Everything after it (fine-tuning) came to $1.27. Teaching a model to behave is cheap; teaching it to know is not.

Figures are actual invoiced Modal usage (GPU + CPU + memory) for this model’s lineage, not estimates. Shared inputs (the QA and preference datasets) are charged at this model’s share, and their generation also used ~$1–2 of external LLM API calls, included above. Serving is billed separately and scales to zero.