Notion Rolls Out Three GPT-5.6 Models for Speed, Cost, and Multi-Day Tasks.

Notion announced on LinkedIn the addition of three GPT-5.6 models to its platform, giving teams new options across performance, cost, and task complexity. The models—Sol, Terra, and Luna—target different workloads, from extended multi-day operations to high-speed responses for lightweight tasks.

GPT-5.6 Sol is designed for long-duration tasks and self-improving agents. Notion positions it for workflows that span days, where agents refine their approach iteratively. GPT-5.6 Terra runs existing GPT-5.5 workloads at roughly half the cost, making it the cost-optimized choice for teams managing volume.

GPT-5.6 Luna focuses on speed for smaller tasks and is currently available for Custom Agents. Custom Agents, recently expanded with worker sharing, allow developers to build task-specific tools that connect to live data and external APIs. Luna's responsiveness fits scenarios where quick turnarounds matter—data lookups, inline suggestions, or rapid prototyping.

The three-model approach mirrors patterns from other platforms offering tiered AI access. Anthropic's Claude and OpenAI's ChatGPT both provide multiple model options, letting users balance cost, latency, and capability depending on the job.

Notion's model expansion follows its Developer Platform launch, which introduced Workers for syncing external data sources, custom agent tooling, and declarative schemas. The platform aims to position Notion as more than a document workspace, turning it into a programmable layer for business workflows.

The company has added several new models over recent months, including Grok 4.5 and Kimi K2.7 Code for in-app coding assistance. The GPT-5.6 additions continue that cadence, offering teams more granular control over how AI fits into their processes.

Key capabilities across the GPT-5.6 models:

• Sol: Multi-day task execution, iterative agent refinement
• Terra: Cost reduction for existing GPT-5.5 workloads
• Luna: High-speed responses for small tasks, available in Custom Agents

Notion has not disclosed pricing details for the new models or whether they will eventually replace older options. Terra's cost positioning suggests Notion is optimizing for teams running high-volume operations where per-query expenses add up.

The three-model structure gives Notion users a practical way to match task type to model performance. Teams building agents that query databases, generate documents, or orchestrate multi-step workflows now have more levers to pull—whether they need extended reasoning, budget efficiency, or sub-second responses.

For developers working with Notion's Custom Agents, Luna's speed advantage could improve real-time interactions. If an agent is surfacing live data or providing inline feedback, latency matters. Terra, meanwhile, offers an immediate path to cost savings for teams already running GPT-5.5-based workflows at scale.

As Notion continues expanding its AI capabilities, the platform is shifting from being a place where teams store information to a layer where that information drives automated work. The GPT-5.6 rollout extends that vision, giving builders more tools to define how AI operates inside their workspace.