TRAINING API
Post-train encoder and decoder models on Fastino's managed infrastructure. Configure your own SFT and RL training jobs, or simply describe your goal to our enterprise agent.
FASTINO TRAINING API
− □ ×
POST /v1/training/jobs
✓ configurable or automated training
✓ tinker compatibility
✓ bring your training loop
✓ managed training for enterprise
✓ encoder / decoder / decision model
› Your model. Fastino’s infra.
FASTINO LABS
POST-TRAIN OPEN WEIGHT MODELS
− □ ×
SELF-SERVICE
WORKFLOW
Your configuration or training loop
YOU PROVIDE
Data and a configuration, or your own training loop
ACCESS
Fastino account
ENTERPRISE
WORKFLOW
Agent-directed training
YOU PROVIDE
A task goal; optional data and evaluation criteria
ACCESS
Enterprise customers
TINKER SDK COMPATIBILITY
Bring your Tinker training loop to Fastino
Keep your data pipeline, reward functions, and evaluation harness. Run supported Tinker SDK training and sampling operations on Fastino by updating the base URL and authenticating with a Fastino key.
Get started
Compare token rates
TINKER SDK
FASTINO
client
Tinker SDK
execution
Fastino Training API
update
Base URL + Fastino credentials
keep
Data, rewards, and evaluations
Customized, open weight models built for finance and healthcare domains, and post-trained on Nemotron 3.5 Super VL using the Fastino Training API outperform base models on various benchmarks.
Fastino-Nemotron-3.5-Super-VL-Finance
New
Executable financial programs, table arithmetic, and numerical extraction: six finance skills in a single rank-32 adapter.
FinQA
35.22→ 77.86%
TAT-QA (F1)
68.53 → 87.51%
SEC-Num
79.21 → 93.88%
ConvFinQA (xfer)
22.33 → 76.33%
Open on Hugging Face
Fastino-Nemotron-3.5-Super-VL-Healthcare
New
Physician-graded conversation, biomedical reasoning, clinical calculation, error correction, and concept extraction: six clinical skills, one adapter.
MedMentions
40.44 → 50.38%
PubMedQA
70.20 → 77.60%
BC5CDR
63.82 → 84.80%
BioRED (xfer)
46.07 → 69.04%
Open on Hugging Face
Post-train supported models with usage-based token pricing.
Token pricing
USD / 1M TOKENS
MODEL
PREFILL
SAMPLE
TRAIN
Nemotron 3.5 Lightning
$0.07 ($0.014 cached)
62% below Tinker ↓
$0.35
29% below Tinker ↓
$0.30
32% below Tinker ↓
Qwen3.6-35B-A3B
$0.08 ($0.016 cached)
82% below Tinker ↓
$0.44
67% below Tinker ↓
$0.44
63% below Tinker ↓
More models coming soon.
Prices are USD per million tokens.
FAQ
Frequently asked questions
