Loading...
News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.
Microsoft's shift to per-token pricing for GitHub Copilot has been dubbed the 'Tokenpocalypse' by Reddit users. On TechCrunch's Equity podcast, hosts discuss how this change signals broader cost pass-through in the AI industry, the rapid rise and fall of tokenmaxxxing, and the challenges AI companies face as they prepare for IPOs amid escalating costs and regulatory uncertainty.
Chinese AI startup DeepSeek announced it will permanently apply a 75% discount on its flagship V4-Pro model, a price cut originally scheduled to expire at the end of May. The move is expected to intensify competition in the global AI market as Chinese firms aggressively court developers with lower pricing.
Deepseek has made its 75 percent discount on the flagship Deepseek V4 Pro model permanent, slashing output token prices to $0.87 per million versus $30 for GPT-5.5. The move intensifies the price war between Chinese and Western AI labs, with Deepseek offering output tokens at roughly 34.5 times less than OpenAI's top model. The permanent pricing, originally set to expire May 31, 2026, also undercuts Anthropic's Opus 4.7 by a similar margin on output tokens.
Models.dev is a new open source project that compiles AI model specifications, pricing, and capabilities into a single community maintained database. The project provides an API for programmatic access, supports SVG provider logos, and encourages contributions via pull requests with TOML based model definitions.
Cerebras Systems completed its initial public offering, raising $5.5 billion after pricing shares at $185, well above the expected range. The stock opened at $385, a 108% jump, and traded above $330 mid-day. The Nvidia rival turned profitable with $510 million in 2025 revenue.
Cerebras completed its initial public offering on Thursday, raising $5.5 billion after pricing shares at $185 each the previous evening. This price exceeded the updated range of $150 to $160 and followed an expansion of the offering to 30 million shares. Pre-market activity shows strong demand from retail investors, pointing to a significant opening jump, with the company achieving a fully diluted valuation of $56.4 billion.
Luma has launched its Uni-1.1 image generation model through a public API with pricing and quality that align with leaders like OpenAI and Google. The REST API supports text-based image creation and editing, ranking 7th or 8th on Arena leaderboards. Two pricing tiers start at $0.0404 per image, with options for reference images and high-resolution outputs.
OpenAI's GPT-5.5 leads the Artificial Analysis Intelligence Index with 60 points, surpassing Claude Opus 4.7 and Gemini 3.1 Pro Preview. Its API pricing shows a net 20 percent increase over GPT-5.4 due to doubled rates offset by 40 percent fewer tokens. The model achieves top accuracy on fact benchmarks yet hallucinates at an 86 percent rate.
Anthropic made a quiet change to its pricing page, moving Claude Code from the $20 Pro plan to the $100 and $200 Max plans. The update sparked backlash on Reddit, Hacker News, and Twitter. Anthropic called it a test for 2% of new signups and soon reverted it amid criticism.
Early measurements show Anthropic's Opus 4.7 model consumes significantly more tokens per request than Opus 4.6, despite identical pricing. Developer Abhishek Ray reported averages like 1.325x for code content and up to 1.47x for technical docs. Community data indicates a 37.4 percent rise in token use and costs across 483 tests, with modest gains in instruction following.
Anthropic launched Claude Opus 4.7, scoring 64.3 percent on SWE-bench Pro, surpassing Opus 4.6's 53.4 percent and OpenAI's GPT-5.4 at 57.7 percent. The model triples image resolution to 2,576 pixels and reduces cybersecurity features during training with new blocks on risky requests. Pricing holds steady per token, but a new tokenizer raises effective costs.