xAI has released Grok 4.5, and the headline number is not a benchmark score but a price tag. According to SiliconANGLE, the new frontier model “dramatically undercuts” Anthropic and OpenAI on cost, positioning Grok’s API as the aggressive value option for developers who compare the AI models database before choosing a provider. For teams already tuning spend against the OpenAI API and Claude, the launch reframes a question that has dominated 2026 procurement conversations: how much is frontier quality really worth per million tokens?
Key takeaways
- xAI has launched Grok 4.5, its newest flagship model, SiliconANGLE reports.
- The outlet says Grok 4.5 “dramatically undercuts” Anthropic and OpenAI on API pricing.
- The release lands as OpenAI rolls out its GPT-Live voice series and prepares a broader GPT-5.6 launch, per SiliconANGLE.
- Anthropic is separately dealing with security-vulnerability claims from Chinese researchers, according to Fox Business.
- Microsoft is reportedly swapping some Anthropic and OpenAI usage for in-house models, Benzinga reports.
- Combined, the moves compress the pricing power that closed frontier vendors have relied on.
What SiliconANGLE reported about Grok 4.5
SiliconANGLE’s report focuses on one central claim: xAI’s newest model, Grok 4.5, undercuts both Anthropic and OpenAI on price, and does so by a dramatic margin rather than a marginal one. The outlet frames the release as a deliberate positioning move against the two labs whose APIs currently anchor most enterprise AI budgets. Specific per-token figures were not included in the snippet available for this report, so buyers evaluating switching costs should consult xAI’s official rate card and re-run their own workloads through an AI API cost calculator before committing.
The signal, however, is unambiguous. When a frontier lab chooses to lead its launch narrative with price rather than capability, it is telling the market that the model is competitive enough on quality to compete primarily on cost — a stance that only makes commercial sense if the underlying inference economics support it.
Why the timing matters for OpenAI API customers
Grok 4.5 arrives in the middle of one of OpenAI’s busiest release cycles of the year. SiliconANGLE separately reports that OpenAI has launched its GPT-Live voice model series ahead of a broader GPT-5.6 rollout, while Politico writes that the company is preparing to release “its most powerful model” after a weekslong internal hold. Fox Business has also reported on OpenAI’s newest model alongside national security plans the company is putting forward.
For customers building on the OpenAI API, this creates a decision window. On one side, OpenAI is layering new capabilities — voice, an upcoming flagship — that may justify premium pricing. On the other, a credible frontier competitor has just walked into the room with what SiliconANGLE describes as a dramatic discount. Procurement teams that had been passively renewing contracts are likely to reopen them.
The competitive backdrop for Anthropic
Anthropic faces its own compounding pressure. Fox Business reports that China has identified security vulnerabilities in Anthropic’s Claude AI model, a story that lands awkwardly for a lab that has marketed safety as a core commercial differentiator. Separately, Benzinga reports that Microsoft — historically one of the largest enterprise buyers of both OpenAI and Anthropic capacity — is swapping some of that usage for in-house AI models.
None of these developments individually threatens Anthropic’s frontier status. Together, they narrow the surface area on which the company can defend premium pricing. If a hyperscaler customer is quietly building alternatives, if a geopolitical rival is publicising vulnerability findings, and if a rival lab is undercutting on cost in the same news week, the pricing conversation becomes harder to hold the line on. Convly’s ongoing AI price-performance index tracks how these shifts move the effective cost-per-quality frontier over time.
How Grok 4.5 fits the 2026 pricing landscape
The Grok 4.5 launch is the most explicit signal yet that frontier AI is entering a genuine price-competitive phase. For most of 2024 and 2025, top-tier closed models commanded a stable premium over open-weights alternatives, on the argument that capability gaps justified the spread. Convly’s open vs closed AI cost study has documented how that gap has been steadily compressing.
A dramatic undercut from another closed frontier lab — as opposed to a cheaper open-weights release — is a different kind of pressure. It attacks the closed-model pricing tier from inside its own category. Buyers who might have been reluctant to switch to a self-hosted stack, and who therefore accepted closed-API premiums, now have a closed-API alternative that removes the price argument for staying put.
| Frontier vendor | Recent development (per cited sources) | Implication for API buyers |
|---|---|---|
| xAI (Grok 4.5) | Launched with pricing that “dramatically undercuts” rivals, per SiliconANGLE | Immediate cost-reduction option worth benchmarking |
| OpenAI | GPT-Live voice series shipped; GPT-5.6 broad release pending, per SiliconANGLE; most powerful model incoming, per Politico | New capabilities may justify premium — or invite a pricing response |
| Anthropic | China reports Claude vulnerabilities, per Fox Business; Microsoft substituting in-house models, per Benzinga | Reputational and demand-side pressure on pricing power |
What developers should evaluate before switching
Price is the loudest signal in this launch, but it is not the only signal that matters for a production migration. Teams considering moving workloads from the OpenAI API or Anthropic to Grok 4.5 will want to test the model on their own task distribution rather than relying on public benchmarks alone. That is especially true for structured-output workloads, tool use, and long-context reasoning, where small quality gaps compound quickly.
For engineering teams building agentic systems, the model choice interacts closely with tooling. Convly’s roundup of AI coding agents shows how differently coding-focused workflows perform across providers, and Grok 4.5’s economics could shift which agent frameworks make sense once real usage is priced in. A sensible migration path is to route a small percentage of traffic through the new model, log quality deltas alongside cost deltas, and only expand once both trend positively.
The broader significance for the AI-model market
Grok 4.5’s pricing posture is significant beyond xAI’s own commercial fortunes because it changes what “reasonable” pricing looks like for a frontier tier. Once a serious lab has publicly anchored the low end, competitors either match, differentiate on capability at their existing price, or lose share on price-sensitive workloads. Historically, in cloud and semiconductor markets, that kind of anchoring has tended to be sticky.
It also strengthens the case for hybrid deployment strategies. Buyers who once viewed frontier APIs as “the expensive option” and self-hosted open-weights as “the cheap option” may increasingly split traffic across multiple frontier APIs plus a self-hosted tier, choosing per-workload based on economics rather than vendor loyalty.
Frequently asked questions
What did xAI announce? xAI launched Grok 4.5, its newest AI model, which SiliconANGLE reports “dramatically undercuts” Anthropic and OpenAI on price.
How much cheaper is Grok 4.5 than the OpenAI API? The exact per-token figures are not included in the source snippet available to this report. SiliconANGLE characterises the undercut as dramatic rather than marginal; buyers should verify current rates directly with each provider.
Does this affect OpenAI’s upcoming launches? Not directly, but it changes the pricing backdrop. SiliconANGLE reports OpenAI is rolling out its GPT-Live voice series ahead of a broader GPT-5.6 release, while Politico reports OpenAI is preparing to release its most powerful model after a weekslong hold.
What is happening with Anthropic? Fox Business reports China has flagged security vulnerabilities in Anthropic’s Claude, and Benzinga reports Microsoft is swapping some Anthropic and OpenAI usage for in-house models — separate developments that together add pressure to Anthropic’s pricing power.
Should teams switch immediately? Not without testing. Price alone rarely justifies migration; teams should benchmark Grok 4.5 on their own workloads for quality, latency, and tool-use behaviour before shifting production traffic.
The bottom line
Grok 4.5’s launch, as reported by SiliconANGLE, is less a capability story than a pricing story — and pricing stories at the frontier have been rare enough in 2026 to be genuinely market-moving. Coming in the same window as OpenAI’s GPT-Live rollout and pending flagship, Anthropic’s vulnerability headlines, and Microsoft’s in-house substitution reports, it marks the moment when frontier AI economics stopped moving in only one direction. For buyers, the actionable response is not to switch reflexively but to rebenchmark: run current workloads through the new model, price them honestly, and let the numbers decide.
Sources: news.google.com. Reported July 09, 2026.
