saltContext Vault
SaltModel Vault
16 / 16 models
Salt-hosted — runs in Salt's secure infrastructure. No data sent to third parties.
External API — bring your own key. You control data sharing with the provider.
Llama 3.3 70B
Meta

State-of-the-art open model. Salt-hosted for maximum data privacy. No data leaves your environment.

Latency: ~110ms
Context: 128k
Privacy: Maximum
Open SourceSalt-hosted
Llama 3.1 8B
Meta

Lightweight, fast inference. Ideal for high-volume, latency-sensitive workflows.

Latency: ~45ms
Context: 128k
Privacy: Maximum
Open SourceFast
Mistral Large 2
Mistral AI

Mistral's flagship model. Excellent reasoning and code. Salt-hosted.

Latency: ~130ms
Context: 128k
Privacy: Maximum
Open SourceSalt-hosted
DeepSeek-R1
DeepSeek

Advanced reasoning model. Salt-hosted. Exceptional on complex analytical and math tasks.

Latency: ~160ms
Context: 64k
Privacy: Maximum
ReasoningSalt-hosted
BioMistral 7B
BioMistral

Fine-tuned on biomedical literature. Optimized for clinical NLP, trial matching, and drug discovery.

Latency: ~60ms
Context: 32k
Privacy: Maximum
MedicalSalt-hosted
Meditron 70B
EPFL

Medical domain LLM trained on clinical guidelines and PubMed. HIPAA-ready deployment.

Latency: ~120ms
Context: 32k
Privacy: Maximum
MedicalClinical
BioGPT
Microsoft

Pre-trained on PubMed abstracts. Ideal for literature review and biomedical Q&A.

Latency: ~55ms
Context: 8k
Privacy: Maximum
BiomedicalPubMed
Legal-BERT
nlpaueb

Pre-trained on legal corpora. Contract analysis, clause extraction, regulatory compliance.

Latency: ~30ms
Context: 16k
Privacy: Maximum
LegalSalt-hosted
FinBERT
ProsusAI

Financial sentiment analysis, earnings call parsing, and regulatory document classification.

Latency: ~30ms
Context: 16k
Privacy: Maximum
FinanceSalt-hosted
Qwen 2.5 72B
Alibaba

Strong performance on financial and multilingual tasks. Salt-hosted.

Latency: ~130ms
Context: 128k
Privacy: Maximum
FinanceMultilingual
Code Llama 34B
Meta

Code generation, review, and debugging. Supports 20+ programming languages.

Latency: ~95ms
Context: 100k
Privacy: Maximum
CodeSalt-hosted
GPT-4o
OpenAI

OpenAI's flagship multimodal model. Bring your own API key. Context is attached via Salt Gateway.

Latency: ~220ms
Context: 128k
Privacy: Standard
External APIBYOK
GPT-4o mini
OpenAI

Cost-efficient OpenAI model. High throughput. Bring your own API key.

Latency: ~90ms
Context: 128k
Privacy: Standard
External APIFast
Claude 3.5 Sonnet
Anthropic

Anthropic's most capable model. Excellent for reasoning, analysis, and long-context tasks.

Latency: ~180ms
Context: 200k
Privacy: Standard
External APIBYOK
Claude 3.5 Haiku
Anthropic

Fast, cost-efficient Claude model. Ideal for high-volume pipelines.

Latency: ~80ms
Context: 200k
Privacy: Standard
External APIFast
Gemini 1.5 Pro
Google

Google's multimodal model with 1M token context window. Bring your own API key.

Latency: ~200ms
Context: 1M
Privacy: Standard
External APIBYOK