Fastino Labs
|
Announcing specialized, open weight variants of Nemotron 3.5 Lightning, developed in collaboration with NVIDIA and trained by the Fastino Fine-Tuning Agent.
Launch
Announcement
Introducing Fastino-Nemotron-3.5-Lightning Finance & Healthcare
Two specialist models built with NVIDIA on Nemotron 3.5 Lightning: open weights, one rank-32 adapter per domain, available today.
Read the announcement
Research
Technical paper
Behind the Fine-Tuning Agent
The founding research behind the Fastino Fine-Tuning Agent's closed-loop training system.
Read the paper
Open weights on Hugging Face · Apache 2.0
Fastino-Nemotron-3.5-Lightning-Finance
New
Executable financial programs, table arithmetic, and numerical extraction: six finance skills in a single rank-32 adapter.
FinQA
15.9→ 59.2%
TAT-QA (F1)
19.0 → 56.6%
SEC-Num
79.7 → 87.6%
Outperforms Nemotron 3 Ultra on BigFinanceBench while activating 3B parameters per token.
Open on Hugging Face
Fastino-Nemotron-3.5-Lightning-Healthcare
New
Physician-graded conversation, biomedical reasoning, clinical calculation, error correction, and concept extraction: six clinical skills, one adapter.
HealthBench
49.7 → 56.2%
PubMedQA
59.0 → 65.0%
MEDEC
50.9 → 63.9%
Within 0.1 points of GPT-5.5 Instant on HealthBench, at a fraction of the active parameters.
Open on Hugging Face
Why Nemotron
Nemotron SOTA in fine-tunability study
In a matched study with identical training data, settings, and evaluations, Nemotron 3.5 Lightning consistently exhibited more fine-tuning improvement than Qwen 3.6
FinQA
+32.6 vs +24.6
TAT-QA exact
+41.5 vs +14.6
BigFinanceBench
+5.6 vs −5.6
Points gained from identical fine-tuning data. Green: Nemotron 3.5 Lightning · grey: Qwen3.6-35B-A3B · orange: regression.
Fine-tune NVIDIA Nemotron on your own data
The autonomous agent that built these models is now available in private preview.
Request access