The Signal — July 10, 2026
The AI price war intensified this week as xAI undercut the field on coding tasks, OpenAI overhauled how voice mode works in ChatGPT, and Illinois staked out some of the toughest AI regulation in the country.
xAI Launches Grok 4.5 — Undercuts Rivals on Price for Coding and Agentic Tasks
xAI released Grok 4.5, a new model purpose-built for coding and agentic workflows, trained in collaboration with Cursor on tens of thousands of Nvidia GB300 GPUs. The headline number: $2 per million input tokens and $6 per million output tokens, undercutting competitors in a market where cost-per-task increasingly matters more than raw benchmark scores.
According to The Decoder, Grok 4.5 uses 4.2 times fewer tokens than Anthropic's Opus 4.8 on comparable tasks, which compounds the pricing advantage. It still trails Fable 5 and GPT-5.5 on standard benchmarks, but xAI is betting that developers building agentic pipelines, where models run thousands of calls per workflow, will optimize for cost over marginal accuracy.
EU availability is expected mid-July. The move puts pressure on every frontier lab to justify premium pricing for coding-focused use cases.
Sources: xAI Official · InfoWorld · The Decoder
OpenAI Ships GPT-Live — Full-Duplex Voice Mode for ChatGPT
OpenAI launched GPT-Live, a full-duplex upgrade to ChatGPT's voice mode that lets the model listen and speak simultaneously, much like a phone call. When a user asks something complex mid-conversation, GPT-Live delegates to GPT-5.5 in the background to generate a more thorough answer without breaking the flow.
GPT-Live-1 is available now for paying ChatGPT subscribers. A lighter "mini" version is free for all users. API access is coming, which will open full-duplex voice to third-party developers building real-time voice agents.
The shift from half-duplex (strict turn-taking) to full-duplex is a meaningful UX change. It removes the awkward pauses that made previous voice modes feel robotic and edges the interaction closer to natural conversation. For OpenAI, it also creates a differentiated consumer experience that is harder to replicate through API wrappers alone.
Sources: OpenAI Official · CNET · The News
Illinois Signed One of the Toughest State-Level AI Laws in the US
Governor JB Pritzker signed SB 315, the Artificial Intelligence Safety Measures Act, into law on July 6. The legislation establishes safety and accountability requirements for AI systems deployed in Illinois, making it one of the most comprehensive state-level AI regulations in the country.
The law joins California's SB 53 as a leading example of states moving ahead of Congress on AI governance. Anthropic CEO Dario Amodei cited both bills as evidence that meaningful AI safety legislation is achievable without stifling innovation. The Illinois act focuses on risk mitigation and transparency obligations for high-impact AI systems.
For AI companies, the practical question is compliance complexity. A growing patchwork of state laws, each with slightly different requirements, is creating the regulatory fragmentation that many in the industry warned about when federal legislation stalled.
Sources: WVIK/NPR · Governor's Office · WGN TV · US News
On the Editor's Desk
We held the GPT-5.6 Sol/Terra/Luna public launch story because the earliest sourcing (Axios, June 26) was two weeks old by the time it rolled out to users on July 9. We also passed on the OpenAI SWE-Bench Pro audit to avoid stacking too much OpenAI coverage in one edition. Several arXiv preprints were considered but lacked sufficient real-world uptake for this edition.
