Snowflake Arctic Instruct
Snowflake's open MoE model: 480B total / 17B active params with dense+MoE hybrid architecture.
Snowflake Arctic Instruct is text & chat AI model from Custom, priced at €0.000 per 1M input tokens with a 4.1K tokens context window.
0.7
Pricing
API Integration
Use our OpenAI-compatible API to integrate Snowflake Arctic Instruct into your application.
npm install railwailimport railwail from "railwail";
const rw = railwail("YOUR_API_KEY");
// Simple — just pass a string
const reply = await rw.run("arctic-instruct", "Hello! What can you do?");
console.log(reply);
// With message history
const reply2 = await rw.run("arctic-instruct", [
{ role: "system", content: "You are a helpful assistant." },
{ role: "user", content: "Explain quantum computing simply." },
]);
console.log(reply2);
// Full response with usage info
const res = await rw.chat("arctic-instruct", [
{ role: "user", content: "Hello!" },
], { temperature: 0.7, max_tokens: 500 });
console.log(res.choices[0].message.content);
console.log(res.usage);Deep dive — Snowflake AI Research's Snowflake Arctic Instruct
Snowflake AI Research is the applied-AI team inside the Snowflake data cloud company, established as a focused unit in 2023 under Yuxiong He (former Microsoft DeepSpeed lead) and Samyam Rajbhandari. Snowflake itself is a public NYSE company (founded 2012) headquartered in Bozeman, Montana with engineering primarily in California. The AI Research team built Arctic to demonstrate that enterprise-targeted LLMs (SQL, code, instruction following) could be trained cheaply by combining many-small-expert MoE designs with carefully curated training data. Arctic Instruct was released April 2024 under Apache 2.0, with the team publicly reporting a training cost of approximately $2M — a fraction of comparable Western frontier runs.
Visit Snowflake AI Research →Snowflake Arctic Instruct is a hybrid dense-MoE transformer released April 2024 under Apache 2.0. The architecture combines a 10B dense transformer with a 128-way MoE component (128 experts of ~3.66B parameters each, top-2 routing) for a total of 480B parameters and 17B active per forward pass. Compared to peer MoEs (Mixtral 8x22B, DBRX) Arctic uses many more, smaller experts and a high expert-count-to-active-param ratio — a deliberate choice to maximise specialisation for enterprise SQL and code tasks at low active compute. Training used a three-stage curriculum on roughly 3.5T tokens of web, code, GitHub, StackExchange and Snowflake-curated enterprise SQL data, with the final stage heavily oversampling SQL and structured outputs. Snowflake reported total training compute of approximately $2M on a cluster of around 3,200 H100 GPUs. Post-training was supervised fine-tuning plus DPO on enterprise instruction data; no full RLHF pipeline was reported.
- Parameters
- 480B total, 17B active per token (10B dense + 128 x 3.66B experts, top-2 routing)
- Context
- 4.1K tokens
- Hybrid 10B-dense + 128-expert MoE architecture
- 17B active parameters out of 480B total — cheap inference for an MoE this large
- Strong text-to-SQL performance (matches or beats Llama 3 70B on Spider, BIRD)
- Solid enterprise code generation in Python, Java, SQL
- Instruction following tuned for structured (JSON) outputs
- Apache 2.0 open weights — fully permissive commercial use
- Reported $2M training cost — milestone in cheap-to-train frontier MoE
- Best for: enterprise SQL generation, structured extraction, self-hosted business assistants.
Pretrained on approximately 3.5 trillion tokens in a three-stage curriculum: broad web data, code-and-math-heavy mix, and a final SQL-and-enterprise-instruction-heavy phase. Sources include filtered Common Crawl, GitHub, StackExchange, books, and Snowflake-curated enterprise SQL corpora. Knowledge cutoff is early 2024. Post-training is supervised fine-tuning plus DPO on enterprise instruction data.
License: Apache 2.0 for both base and Instruct weights. Commercial use, redistribution and modification permitted without royalty.
Known limitations
- Very short 4K context window for a 2024 release
- Weaker on open-ended creative writing than Llama 3 70B Instruct
- Multilingual support limited — trained primarily on English
- Reasoning chains shorter than dedicated reasoning models
- No vision or audio input
Frequently asked questions
Related Models
View all Text & ChatBio_ClinicalBERT
The original Bio_ClinicalBERT from Alsentzer et al., a BERT model initialized from BioBERT and further pretrained on all MIMIC-III clinical notes. Served as a fill-mask endpoint it predicts masked tokens in clinical text and produces clinical embeddings. It is the standard encoder backbone behind many downstream clinical NLP fine-tunes.
Biomedical NER (all entities)
Token-classification model from d4data that tags 84 biomedical entity types in clinical and medical text, including disease, sign, symptom, medication, dosage, lab value, body part and procedure. Trained on the Maccrobat clinical case corpus on a DistilBERT base, so it runs cheaply for high-volume tagging.
Claude Opus 4
Anthropic's most powerful model. Exceptional at complex analysis, agentic tasks, and extended reasoning.
Claude Opus 4.8
Anthropic's most capable Opus-tier model. State of the art on long-horizon agentic work, coding and knowledge tasks, with a 1M-token context window at standard pricing.
Start using Snowflake Arctic Instruct today
Get started with free credits. No credit card required. Access Snowflake Arctic Instruct and 100+ other models through a single API.