AI This Week: The Model Price War Is Here
Two competing AI model families launched within 24 hours of each other this week, putting real downward pressure on API pricing for the first time this year. Meanwhile, Microsoft reversed course on one of its most aggressive AI feature rollouts after a user revolt. Here’s what each development means for your business.
OpenAI Launched a Three-Tier Model Family — and It’s Already Running Inside Microsoft 365
On July 9, OpenAI made its GPT-5.6 family generally available across ChatGPT, its Codex platform, and the API, per TechCrunch’s launch coverage. The family has three named tiers: Sol is the flagship at $5 per million input tokens and $30 per million output tokens; Terra is the midrange option at $2.50 input / $15 output; and Luna is the high-volume budget tier at $1 input / $6 output.
The same day, OpenAI announced on its own site that GPT-5.6 is now the preferred model in Microsoft 365 Copilot — meaning it runs inside Word, Excel, PowerPoint, and Teams Chat by default for licensed users, per the Microsoft 365 Copilot blog. If your business pays for Microsoft 365 Copilot, you are already running the new model family.
The tiered structure removes a real friction point in AI API decisions. For the last two years, businesses either paid flagship rates or stepped down to a meaningfully weaker model. Sol handles the most complex reasoning and research tasks. Luna handles high-volume, lower-stakes work at a fraction of the cost. Terra lands in the middle — capable enough for most business workflows, priced for everyday use.
What this means for your business: Terra at $2.50 per million input tokens is where most SMB evaluation should start. It undercuts Anthropic’s Claude Sonnet 5 post-August-31 rate of $3 per million input tokens and comes with OpenAI’s established enterprise tooling. If you run any AI-assisted document workflow — proposal drafts, invoice processing, customer email replies — benchmark Terra against your current model before renewing your vendor contract. For businesses in wholesale distribution or construction processing high volumes of similar documents, Luna’s $1 rate for bulk classification tasks is worth a separate look.
SpaceXAI Entered the Price War with Grok 4.5 — One Day Earlier
On July 8, SpaceXAI released Grok 4.5, scooped by Axios and covered by TechCrunch. The pricing is $2 per million input tokens and $6 per million output tokens — positioning it below OpenAI’s Luna tier on output cost and below Terra on both input and output for most workflows.
Elon Musk described the model as “an Opus-class model, but faster, more token-efficient, and lower cost.” The model is purpose-built for coding and multi-step agentic tasks — AI systems that take sequences of actions rather than producing a single answer — and launched inside Cursor (an AI coding tool) and the SpaceXAI API console. EU availability is expected by mid-July.
The timing was not accidental. Two of the largest AI providers put meaningfully competitive products on the market within 24 hours of each other, both priced to undercut the current market leader on at least one dimension. That dynamic — providers competing on price rather than just capability benchmarks — is new for the frontier AI market and directly benefits businesses buying API access.
What this means for your business: If you are paying current frontier-model rates to power document processing, content generation, or automated workflows, this week’s launches are a genuine signal to re-evaluate your cost model. Grok 4.5’s pricing is attractive for agentic and coding use cases, but it is a new release from a provider with a shorter enterprise track record than OpenAI or Anthropic. Benchmark it in a test environment before migrating any live workflow. The broader takeaway is the one worth holding onto: the price of capable AI is dropping, and the competitive dynamics now favor buyers who are willing to evaluate more than one provider.
Microsoft Teams Backed Down on Ambient AI — and It Should Make You Check Your Settings
The third story this week is a reversal rather than a launch. After months of rolling out AI features inside Teams, Microsoft added a mid-meeting toggle that lets licensed meeting organizers turn off Copilot, Facilitator, and Intelligent Recap during live sessions, per TechRadar’s reporting and Windows Central’s coverage. Targeted Release rollout began the first week of July with General Availability expected through the end of the month.
The reversal followed user backlash to Facilitator, an ambient AI assistant that requires being activated to monitor and transcribe meetings in real time. Teams had been deploying AI features more aggressively with each update cycle, and both end users and IT administrators raised concerns about whether meeting participants always understood they were being observed by AI. The new toggle lets organizers disable all AI features at once or turn off Copilot, Facilitator, and Recap independently.
This is a pattern worth recognizing. Enterprise software vendors pushed AI on by default, discovered that consent and privacy concerns surface quickly when employees and external participants aren’t informed, and then pulled back. The Teams backlash happened at large enterprise scale — but the same dynamic plays out at smaller companies where the stakes feel more personal.
What this means for your business: If you run Microsoft 365 with Teams Premium or Copilot licensing, log into your admin settings and verify whether Facilitator and Intelligent Recap are on by default in your tenant. Businesses in specialty contracting and freight and logistics regularly bring clients, vendors, and subcontractors into Teams calls. Those participants may not know an AI is transcribing and summarizing the meeting. Getting ahead of that conversation — a one-paragraph policy stating which meetings use AI transcription and which don’t — takes fifteen minutes and prevents a harder conversation later.
The Takeaway
Three developments in one week tell a consistent story: capable AI is getting cheaper, provider competition is real, and the era of AI features deployed without user awareness is facing resistance. For your business, both the pricing news and the Teams rollback point in the same direction. If you haven’t compared AI vendors or API costs recently, this week gave you two new pricing reference points. If you’ve deployed AI tools inside your team without a clear communication to staff about what they do, the Teams backlash is a useful preview of what happens when that gets skipped.
Getting AI working inside a business takes more than picking the cheapest model. If you want help working through the options — what to evaluate, what to build first, how to get your team on board — BlueHill is here. We work with Houston SMBs to turn AI capability into measurable business results.