On July 9, OpenAI moved GPT-5.6 from limited government-vetted preview to full public release, retiring the old "mini/nano" naming in favour of three durable capability tiers. Sol is the flagship for hard, long-running reasoning; Terra matches GPT-5.5 at half the cost; Luna is the fast, cheap workhorse — priced at $5/$30, $2.50/$15 and $1/$6 per million tokens respectively.
Sam Altman told CNBC that Sol is 54% more token-efficient on agentic coding, and OpenAI is pairing the launch with ChatGPT Work — an agent that folds Codex into a single desktop app to handle documents, spreadsheets and multi-step business tasks with minimal oversight. The rollout followed a White House-linked review process; a White House official told CNBC the administration gave no formal "green light," leaving release decisions with OpenAI itself.
SK Hynix priced 177.9 million ADRs at $149 each on Thursday, surpassing Alibaba's 2014 record to become the largest U.S. share sale ever by a foreign company. Shares opened Friday at roughly $170–$171 on the provisional ticker SKHYV, a gain of about 14–15%, pushing market cap above $1.2 trillion. Demand reportedly outstripped supply more than seven to one.
SK Hynix holds around 56% of the global high-bandwidth memory market and supplies Nvidia directly — the listing gives U.S. investors their first direct route into the world's dominant HBM producer, and analysts expect it to narrow the valuation gap with Micron.
Muse Spark 1.1 launched in U.S. public preview at $1.25 input / $4.25 output per million tokens — well below GPT-5.5's $5/$30 and Claude Opus 4.8's $5/$25. It's a reasoning model tuned for repository-level coding, computer-use tasks, and multi-agent orchestration, with a 1M-token context window and a $20 free-credit onboarding tier for new developers.
Analysts told InfoWorld the pricing is a "shot across the bow" that could force OpenAI and Anthropic to compete harder on inference economics, even as some expect Meta to raise prices 30–50% once market share solidifies.
Ollama closed a $65 million Series B led by Theory Ventures, bringing total funding to $88 million. Founder Jeff Morgan says the tool — which lets developers run open-weight models locally in one command — now reaches 8.9 million developers monthly, up from half that in January, adding close to a million installs a week.
Benchmark's Peter Fenton, who joined Ollama's board after leading its Series A, argues open-weight models will generate the "supermajority" of AI tokens within 18–24 months — framing Ollama as the platform layer everything else plugs into.