What is the largest context window available with Atlas and Snowflake Cortex?

The largest context window available through Atlas with Snowflake Cortex is 1,050,000 tokens, provided by GPT-5.5. Claude Opus 4.8, Claude Fable 5, and Claude Sonnet 4.6 each offer a 1,000,000 token context window.

How is Snowflake Cortex priced when used with Atlas?

Snowflake Cortex models are billed in Snowflake credits, not on a per-Mtok basis. Pricing is not publicly listed, and credit consumption must be measured after an Atlas agent run.

Can I use Snowflake Cortex with Atlas if I don't have a Snowflake account?

No, a Snowflake account is a non-negotiable prerequisite for using Snowflake Cortex with Atlas. The base URL for model access is templated on your Snowflake account identifier.

What security benefits does Snowflake Cortex offer for Atlas users?

Snowflake Cortex places frontier models within your Snowflake security boundary. Model access is governed by the same RBAC (Role Based Access Control) that manages your Snowflake warehouses, ensuring consistent and robust security.

Which frontier models are available through Snowflake Cortex for Atlas?

Snowflake Cortex spans a range of frontier models, including OpenAI (GPT-5 through GPT-5.5), Anthropic (Haiku 4.5 through Opus 4.8, Claude Fable 5), Google (Gemini 3.1 Pro), Mistral Large, and DeepSeek-R1, all from one endpoint.

How do I configure Atlas to use a specific Snowflake Cortex model?

After exporting your `SNOWFLAKE_ACCOUNT` and `SNOWFLAKE_CORTEX_PAT`, you can pin a model like `"snowflake-cortex/claude-opus-4-8"` in your `atlas.json` configuration file. You can also set a `small_model` for efficiency.

Is it possible to estimate the cost of an Atlas agent run using Snowflake Cortex before it executes?

No, it is not possible to estimate the cost of an Atlas agent run using Snowflake Cortex before execution. The registry carries no per-token cost, and credit consumption must be measured after the fact.

Atlas with Snowflake Cortex in 2026

Atlas with Snowflake Cortex provides a direct, secure pathway for developers in 2026 to leverage frontier AI models, including Claude Opus 4.8 and GPT-5.5, directly within their Snowflake environment. This integration is ideal when your code and data reside in Snowflake, offering a 1,000,000 token context window for top Anthropic models. Pricing is not publicly listed, as it is billed in Snowflake credits, not per-Mtok.

Why choose Snowflake Cortex for Atlas?

In 2026, Atlas users seeking to integrate their AI coding agent directly with Snowflake data and code will find Snowflake Cortex an optimal choice. It provides the shortest legal path from an Atlas agent to a production table, leveraging models like Claude Opus 4.8 with a 1,000,000 token context.

Snowflake Cortex is designed for developers whose code and data ecosystems are already within Snowflake. It places frontier models, including Anthropic's Claude Opus 4.8 and Claude Fable 5, both offering a 1,000,000 token context window and 128,000 max output, directly behind your Snowflake account. This means model access is governed by the same robust RBAC that secures your Snowflake warehouses, ensuring a consistent security posture. Atlas connects to Model Context Protocol servers, exposing their tools to the agent, making this integration direct for tasks requiring secure data access.

What are the context window and pricing for Snowflake Cortex models?

Snowflake Cortex offers substantial context windows, with GPT-5.5 providing 1,050,000 tokens and top Anthropic models like Claude Opus 4.8 and Claude Fable 5 offering 1,000,000 tokens. Pricing for Snowflake Cortex models is not publicly listed in 2026, as it is billed in Snowflake credits rather than per-Mtok.

The context window for models available through Snowflake Cortex varies by model. GPT-5.5 offers a 1,050,000 token context window, while Claude Opus 4.8, Claude Fable 5, and Claude Sonnet 4.6 each provide a 1,000,000 token context window. This allows Atlas to process extensive codebases and documentation. Regarding cost, Snowflake Cortex models are billed in Snowflake credits, not on a per-token basis. This means that while there is no per-token cost listed, the actual credit consumption for an Atlas agent run must be measured after the fact, making upfront cost estimation challenging.

What are the tradeoffs of using Snowflake Cortex with Atlas?

While Snowflake Cortex offers robust security and access to frontier models like Claude Opus 4.8, developers in 2026 must consider its specific tradeoffs. A primary concern is that credit consumption for agent runs cannot be estimated from the model list, requiring post-fact measurement.

The main tradeoff for Atlas users considering Snowflake Cortex is the pricing model. Since the registry carries no per-token cost, estimating the price of an agent run from the model list is not possible. Credit consumption must be measured after the fact, which can complicate budgeting and cost optimization efforts. Furthermore, Snowflake Cortex is entirely dependent on having an active Snowflake account. The base URL for access is templated on your account identifier, making it useless without this prerequisite. Atlas allows switching the active model and provider on the fly, so users can easily experiment with other models if these tradeoffs are not suitable for a particular project.

When should I choose a different model for Atlas?

Developers in 2026 should consider alternative models for Atlas if they lack a Snowflake account or require transparent, per-token pricing for immediate cost estimation. If your data and code do not reside within Snowflake, the core benefit of Snowflake Cortex is diminished.

If you do not have a Snowflake account, Snowflake Cortex is not a viable option, as its access is intrinsically linked to your Snowflake account identifier. Additionally, if your primary concern is precise, upfront cost estimation based on per-token usage, Snowflake Cortex's credit-based billing model, which requires post-fact measurement, might not be ideal. Atlas lets you switch the active model and provider on the fly, offering flexibility to choose models with publicly listed per-token pricing or those that integrate better with non-Snowflake data environments. This ensures you can always select the best tool for the specific task and budget.

Setup

01Export your Snowflake account identifier: `export SNOWFLAKE_ACCOUNT="your_account_identifier"`
02Export your Snowflake Cortex programmatic access token: `export SNOWFLAKE_CORTEX_PAT="your_pat_token"`
03Confirm the catalog resolves against your account host: `atlas models snowflake-cortex`
04Pin a frontier model in `atlas.json`: `{"model": "snowflake-cortex/claude-opus-4-8"}`
05Alternatively, pin a cheaper 1M-context slot: `{"model": "snowflake-cortex/claude-sonnet-4-6"}`
06Set a smaller model for efficiency: `{"small_model": "snowflake-cortex/openai-gpt-5-nano"}`

Frequently asked questions

What is the largest context window available with Atlas and Snowflake Cortex?: The largest context window available through Atlas with Snowflake Cortex is 1,050,000 tokens, provided by GPT-5.5. Claude Opus 4.8, Claude Fable 5, and Claude Sonnet 4.6 each offer a 1,000,000 token context window.
How is Snowflake Cortex priced when used with Atlas?: Snowflake Cortex models are billed in Snowflake credits, not on a per-Mtok basis. Pricing is not publicly listed, and credit consumption must be measured after an Atlas agent run.
Can I use Snowflake Cortex with Atlas if I don't have a Snowflake account?: No, a Snowflake account is a non-negotiable prerequisite for using Snowflake Cortex with Atlas. The base URL for model access is templated on your Snowflake account identifier.
What security benefits does Snowflake Cortex offer for Atlas users?: Snowflake Cortex places frontier models within your Snowflake security boundary. Model access is governed by the same RBAC (Role Based Access Control) that manages your Snowflake warehouses, ensuring consistent and robust security.
Which frontier models are available through Snowflake Cortex for Atlas?: Snowflake Cortex spans a range of frontier models, including OpenAI (GPT-5 through GPT-5.5), Anthropic (Haiku 4.5 through Opus 4.8, Claude Fable 5), Google (Gemini 3.1 Pro), Mistral Large, and DeepSeek-R1, all from one endpoint.
How do I configure Atlas to use a specific Snowflake Cortex model?: After exporting your `SNOWFLAKE_ACCOUNT` and `SNOWFLAKE_CORTEX_PAT`, you can pin a model like `"snowflake-cortex/claude-opus-4-8"` in your `atlas.json` configuration file. You can also set a `small_model` for efficiency.
Is it possible to estimate the cost of an Atlas agent run using Snowflake Cortex before it executes?: No, it is not possible to estimate the cost of an Atlas agent run using Snowflake Cortex before execution. The registry carries no per-token cost, and credit consumption must be measured after the fact.

Try SeaShell in your terminal

The terminal-native AI coding agent. Free core, single binary.

Install SeaShell

Related guides

Atlas for Angular in 2026

Adopt Atlas, the terminal-native AI coding agent, for your Angular projects in 2026. Enhance development with intelligent code search, secure local embeddings, and granular control over AI actions.

Review a Pull Request with Atlas (2026 Workflow)

How to review a pull request with Atlas in 2026: bash produces the raw patch, read pulls whole files, the lsp tool's findReferences checks callers the diff never shows.

Migrate a Deprecated API Across Every Callsite with Atlas (2026 Workflow)

How to migrate a deprecated API across every callsite with Atlas in 2026: the lsp tool's findReferences enumerates callers, todowrite tracks them, apply_patch migrates each one.

Atlas for SQL in 2026

Atlas is a terminal-native AI coding agent for SQL in 2026. Run it in a repo with your migrations or .sql files, optimize a query, and review the diff before applying.

Atlas for Pandas: Terminal-Native AI Coding in 2026

Atlas is a terminal-native AI coding agent for Pandas. Vectorize df.apply, fix chained assignment under Copy-on-Write, and pin DataFrames with assert_frame_equal.

Atlas vs. Graphite: Choosing Your AI Coding Agent in 2026

Atlas and Graphite comparison for 2026. Atlas offers terminal-native AI with permission-gated tools and local embeddings. Graphite provides stacked PRs and AI review for GitHub teams.

Atlas for Astro: Islands, Content Collections, and Zero JS by Default in 2026

Atlas is a terminal-native AI coding agent for Astro in 2026. It reads astro.config.mjs, src/pages, and content collection schemas, drops needless client:load directives, and runs astro check.

Atlas for C# in 2026

Atlas is a terminal-native AI coding agent for C# and the .NET SDK in 2026. Run it in a solution with a .csproj or .sln and approve every diff before dotnet build.

Browse this resource hub