AI Provider Cost Updates 2026-11-06 | AI Economics News
OpenAI
Pro and Business Premium plans now include more Work and Codex usage
OpenAI just bumped up the usage included in its Pro and Business Premium plans, and the change is a big one for anyone planning AI spend.
The company increased the Work and Codex usage that comes with these plans. Pro 100 and Pro 500 now get twice their previous allowance.
If you’re budgeting for team workloads on these plans, you now get more included usage than before, so it’s worth revisiting your usage forecasts and plan choices.
GPT-6 Rosalind Research is live for approved life sciences users
OpenAI has released GPT-6 Rosalind Research, a new model for approved life sciences users.
It’s available in Work, Codex, and the API. That said, the rollout is limited for now to a subset of Enterprise and Business customers.
If your organization is eligible, this new model may change how you think about workload economics and model selection. It’s a good moment to check with your teams about whether they have access.
DevRev shows how GPT-6.1 Sol balances reasoning, speed, and cost
Looking for a real-world example of picking a model with cost in mind? OpenAI shared how DevRev uses GPT-6.1 Sol.
DevRev uses the model to power its Computer and Operator products. OpenAI highlights the balance of reasoning, speed, and cost that matters for enterprise service workflows.
Because the model is positioned around cost and performance efficiency for production use, it’s a handy reference point if you’re weighing similar choices for your own deployments.
Images 2.6 brings faster generation and new API controls
OpenAI introduced Images 2.6, an update to its image tools across ChatGPT, Codex, and the API.
The release offers faster image generation, GIF creation, and more precise editing. It also adds new API controls.
Faster generation and tighter controls can help image-heavy workflows run more efficiently and cut down on waste, which is good news for teams watching their image-related spend.
GPT-5.6 Cyber is now available in AWS GovCloud (US-East) through Amazon Bedrock
OpenAI made GPT-5.6 Cyber available in AWS GovCloud (US-East), announced as Daybreak Red availability.
Access comes through Amazon Bedrock and is open to approved customers. Even though it’s region-specific, it gives enterprises more choices for where they deploy security workloads.
If you manage costs for teams in regulated environments, this gives you a new deployment option to factor into planning.
Dots is generally available for Enterprise, Edu, and Healthcare workspaces
Dots is now generally available for eligible Enterprise, Edu, and Healthcare workspaces.
The feature brings persistent task continuity, so work can carry on across sessions. It also includes admin controls and usage review.
On top of that, it supports data residency and Enterprise Key Management. Together, these can help large deployments stay organized and well governed.
Heimstaden uses ChatGPT Enterprise and Codex to cut property management costs
OpenAI shared a customer story about Heimstaden, a company working to change residential real estate management with AI.
According to OpenAI, Heimstaden uses ChatGPT Enterprise and Codex to transform property management and cut costs across Europe.
It’s not a pricing change, but it’s a useful large-scale example of an enterprise using AI to improve efficiency and reduce costs.
Anthropic
Claude apps gateway adds a cost analytics API and richer usage data
Anthropic’s Claude Code v2.1.317 is a big win for cost reporting, especially if you handle allocation and chargeback.
The Claude apps gateway now exposes a cost analytics API. You can pull spend by date range, developer, day, or identity provider (IdP) group.
The release also adds new OpenTelemetry (OTel) attributes for plan, seat type, provider, and custom base URL. With that extra detail, it’s easier to analyze cost and usage across your teams.
Claude Code can now attribute AWS costs per developer and run cheaper side jobs
Claude Code v2.1.318 makes per-developer cost tracking on Amazon Web Services (AWS) much simpler, which matters for FinOps and chargeback.
The release adds assume_role support for per-developer AWS cost attribution, so you can see who’s driving spend.
It also adds an effort option for model forks. Low-effort side jobs can then run more cheaply instead of using the same resources as your main task.
Finally, several reliability fixes cut down on wasted retries and stuck sessions, so you spend less on work that doesn’t finish.
Pooled group budgets arrive in the Claude apps gateway
Claude Code v2.1.316 gives you a new way to keep spend in check with pooled group budgets in the Claude apps gateway.
These budgets set shared monthly limits for a group, and they sit on top of per-user limits. That means you can control spend at both the individual and the group level.
The release also adds idle compaction for long conversations. It can reduce token consumption and improve workload efficiency across supported providers.
Message threads help long-running agents send less data on every turn
Anthropic introduced message threads in the Messages API, and they could make long-running agent workflows a lot leaner.
With message threads, clients send the conversation once. After that, each turn only needs to submit the incremental changes.
Smaller prompt payloads can materially reduce token spend, which adds up quickly for agents that run for a long time.
Gemini Enterprise now supports prepaid credits for Pay-as-you-go billing
Google Cloud made prepaid credits generally available for Gemini Enterprise self-serve billing accounts, giving you more control over how you pay.
Customers can buy credits and set auto-reload with spend limits. The credits cover Pay-as-you-go usage as well as usage above quota.
They work across the Gemini Enterprise app, Agent Platform, and AI developer tools. That means better billing control and easier cost monitoring in one place.
Gemini Enterprise adds a Salesforce federated data store
Gemini Enterprise now supports a Salesforce federated data store and it’s generally available.
Users can query and act on Salesforce data without ingesting it first. This can reduce duplicated data movement and storage overhead.
It also helps workflow efficiency for enterprise search and automation use cases.
