OpenAI chip and Anthropic allegations define a busy AI day

OpenAI and Broadcom unveiled Jalapeño on June 24, 2026, positioning the custom inference chip as a way to cut LLM serving costs. Early lab tests showed approximately 50% lower inference cost per token than current-generation Nvidia GPUs, while broader deployment is expected to move from prototypes by end of 2026 to production ramp in 2027 and full scale in the first half of 2028.

Anthropic accused operators affiliated with Alibaba and Qwen of using approximately 25,000 fraudulent accounts to generate more than 28.8 million exchanges with Claude between April 22 and June 5, 2026. Google DeepMind faced four senior departures in six days, with Jonas Adler and Alexander Pritzel joining Anthropic after Noam Shazeer moved to OpenAI and John Jumper joined Anthropic.

GPT-5.6 drew prediction-market attention, with Polymarket contracts exceeding $1.1 million priced at 83% probability for a launch before June 28, 2026. AI economics also shifted, as DeepSeek made V4-Pro pricing permanent at $0.44 per million input tokens and $0.87 per million output tokens, while the AI code tools market was estimated at $9.3 billion in 2026.

Apple introduced AFM 3 foundation models, Google delayed Gemini 3.5 Pro to July, and Gemini 3.5 Live Translate began rolling out with support for more than 70 languages. Cognition’s FrontierCode benchmark put Fable 5 ahead on production-quality pull requests at 46.3%, compared with Claude Opus 4.8 at 34.3% and GPT-5.5 at 25.5%.