Benchmark claim
Google DeepMind says Gemini 3.0 Ultra tops several reasoning benchmark suites
Gemini benchmark wins help narrative momentum even before market share shifts.
Major model release
xAI begins closed beta for Grok 5 after latest Colossus training cycle
Grok 5 beta is xAI's strongest bid yet to be treated as a serious frontier lab.
Major model release
DeepSeek releases R2 reasoning model under an open-weight license
DeepSeek R2 reignites the open-weight reasoning debate with strong developer attention.
Major model release
Meta publishes Llama 5 400B open-weight checkpoint for research partners
Llama 5 400B keeps Meta central to the open-weight conversation.
Infrastructure expansion
NVIDIA says Blackwell Ultra racks have started shipping to top hyperscalers
Actual Blackwell Ultra shipments are one of the strongest possible infrastructure signals.
Major product launch
Mistral launches Mistral Large 3 with tighter controls for enterprise buyers
Mistral Large 3 is a needed proof point that the company can keep shipping for enterprise buyers.
Controversy
Enterprise CIO survey points to rising Copilot seat churn at large Microsoft deployments
Copilot churn reporting hits directly at Microsoft's AI monetization story.
Scores decay over time, with roughly 10% of an event's impact fading each day. `score_change_24h` is the current score minus the score 24 hours ago, and `score_change_7d` is the current score minus the score 7 days ago. Every score change is tied to a specific recent event rather than a black-box ranking rule.