GPT-5.6 debuts, prices race down, and schools go AI-native
OpenAI previews GPT-5.6 in three tiers, DeepSeek locks in steep cuts, and Microsoft finds 92% of students already use AI.
By BINA Editorial
Two model launches, a permanent price-war salvo, and a global education survey converge on a single message: AI is no longer arriving — it is already here.
GPT-5.6 arrives in limited preview, three tiers of capability
OpenAI released a limited preview of GPT-5.6 on June 26, offering three model tiers named Sol, Terra, and Luna. Sol targets complex, long-horizon tasks; Terra is designed for everyday professional use; Luna optimises for cost-efficient large-scale calls. Access remains restricted to a small group of partner institutions while OpenAI monitors performance before a wider rollout. The launch completes a months-long preview cycle — prediction markets had priced an 83 percent probability of release before June 28. A context window rising to 1.5 million tokens (up from 1 million in GPT-5.5) and faster Codex response times are among the reported improvements.
DeepSeek makes its 75% price cut permanent
Chinese AI laboratory DeepSeek announced that the 75 percent promotional discount it applied to its flagship V4-Pro model will not expire: the new rates are now permanent. V4-Pro now costs $0.44 per million input tokens and $0.87 per million output tokens, compared with $2.50 and $15.00 for OpenAI's GPT-5.5. Cache-hit pricing — critical for AI agents that send repeated context — has in some configurations fallen to one-tenth of previous costs. The company attributes its ability to sustain these prices to the V4 series being optimised to run on Huawei's Ascend AI accelerators rather than Nvidia chips, insulating DeepSeek from US export restrictions. The move widens the price gap between Chinese and American frontier models and puts pressure on competing labs to reconsider their own rate cards.
Microsoft: 92% of students use AI, but 77% have had no formal training
Microsoft published the third edition of its AI in Education Report on June 24, drawing on surveys of students, educators, and school leaders globally. Ninety-two percent of students and education leaders, and 88 percent of educators, reported using AI for school-related purposes. Despite that reach, 77 percent of students and 53 percent of educators have received no formal AI training — the central concern the report flags for institutions worldwide. Microsoft announced new no-cost tools for schools: AI-assisted unit plans, student AI guidelines, a Study and Learn Agent, and a free AI Literacy for Educators credential pathway through its Elevate for Educators programme. The findings echo a pattern visible across other sectors: adoption has outpaced preparation.
Apple rebuilds Siri from the ground up at WWDC 2026
At its annual developer conference, Apple introduced Siri AI, a rebuilt version of its assistant described as significantly more capable and context-aware than its predecessor. Unlike the earlier Siri, the new version takes actions across multiple apps in response to what users are doing in the moment — adding a song to a playlist mid-conversation, editing a recently sent message, or surfacing a confirmation code from Mail when a user calls an airline. Apple Intelligence features shipping with iOS 27 also include AI-powered photo editing tools, natural-language Shortcuts automation, and calendar creation from conversational input. The assistant carries context across devices, so a question started on iPhone can be continued on iPad. Availability is set for autumn 2026 in supported languages across compatible hardware.
EU finalises Code of Practice on AI-generated content labelling
The European Commission published its final Code of Practice on the marking and labelling of AI-generated content, a key transparency measure under the EU AI Act. Signatories — drawn from major platform operators — commit to visible, machine-readable labels on synthetic images, audio, and video distributed through their services. The code is voluntary but carries strong expectation of broad adoption ahead of the AI Act's transparency provisions entering into force in August 2026. Separately, the Commission closed a public consultation on draft guidelines for classifying high-risk AI systems on June 23; a finalised version is expected before year-end, ahead of the December 2027 compliance deadline for stand-alone high-risk systems.
Google makes Gemini 3.5 Flash the default for Search AI Mode globally
Google rolled out Gemini 3.5 Flash as the default model powering Search AI Mode, extending the upgrade to all users worldwide. The company described the change alongside what it called the most significant redesign of the Google Search interface in over 25 years, centred on an AI-powered search box that supports sustained conversational queries. Gemini 3.5 Flash debuted at Google I/O in May at $1.50 per million input tokens and $9.00 per million output tokens, and has been the default in the standalone Gemini app since then. Its deployment as the backbone of Search AI Mode signals that Google considers it capable enough for high-volume, latency-sensitive consumer use — a concrete step in the company's shift from keyword retrieval toward agent-driven answers at scale.