For organizations building custom Copilot Studio agents or integrating Work IQ into Power Platform workflows, Work IQ API access is a separate line item from the per-user Copilot license. Budget accordingly — Work IQ API usage is what powers the "grounded in your data" claim that custom agents make, and the credit cost scales with the depth of grounding.
GPT-6 Astra in Microsoft Foundry
GPT-6 Astra is generally available in Microsoft Foundry since September 3, 2026. Astra is offered with Standard and Provisioned Throughput deployment options in both Global and US Data Zone geographies. Standard provides pay-as-you-go flexibility for variable demand, while Provisioned Throughput provides dedicated model-processing capacity for workloads requiring consistent latency and guaranteed throughput (Microsoft Azure blog, September 3, 2026).
Astra availability through Foundry is part of a broader pattern: Microsoft has been expanding model choice in Foundry since adding OpenAI as a subprocessor in July 2026. Foundry customers can select from OpenAI models operated by Microsoft (Azure OpenAI), OpenAI-operated models through OpenAI as subprocessor, and direct Azure models from Cohere, DeepSeek, Meta, Mistral AI, and xAI. The September 2026 Copilot Credits Guide formalizes the billing path for all of these (Microsoft Foundry, September 4, 2026).
What enterprise buyers should do
Three actions for organizations deploying Microsoft Copilot at scale.
- Map current Copilot workloads to credit consumption patterns. Chats consume few credits; Cowork, Studio agents, and Dynamics 365 AI consume many. The pricing model rewards shifting stable, predictable workloads to the per-user subscription and isolating bursty agent workloads to credit consumption.
- Provision the Copilot Credit Pre-Purchase Plan (P3) via MACC. This gives committed-use pricing for credit consumption and is the recommended procurement path for enterprises with stable Azure commitments.
- Plan Work IQ API access as a separate budget line. Custom agents and AI solutions using Work IQ APIs are billed separately from the per-user subscription. The 80% token-reduction figure applies to agent execution but does not change the billing model.
What to watch next
Three near-term datapoints. First, the October 2026 Microsoft Ignite conference will surface the next set of Copilot Credits pricing details and workload expansions. Second, the rollout of Copilot Cowork into general availability beyond enterprise tenants will change the credit-consumption profile materially. Third, any expansion of the per-user Copilot subscription to include more Copilot Credits by default would shift the breakeven point between per-seat and usage-based pricing for many enterprises.