The latest in models from spAIsee.
Google DeepMind’s AlphaEvolve uses Gemini models, automated evaluators and evolutionary search to discover algorithms, revealing why measurable objectives, independent validation and human release gates matter.
Microsoft CEO Satya Nadella says AI agents need independent permissions, audit logs and a human-controlled emergency brake as they gain access to tools and business systems.
PrismML’s 5.9 GB Bonsai 2 27B brings compressed AI closer to laptops and phones, promising private offline assistants while independent tests question its benchmark performance, compatibility and reliability.
Claude Opus 5.5 may have reduced familiar AI writing tells, but Graphite’s research suggests new fingerprints are emerging in its vocabulary, punctuation and communication style.
Musubi’s open-weight PolicyLM-1.7B aims to make content moderation faster, cheaper and easier to update by translating written rules into constrained decisions, though real-world reliability remains unproven.
Anthropic’s GRAM research could let AI companies package sensitive capabilities as removable modules, offering finer access controls than refusals while raising difficult questions about isolation, scaling, fine-tuning and security.
Mistral’s Le Chonk, a trillion-parameter multimodal model, puts Europe’s open AI ambitions to the test as questions mount over benchmarks, safety, operating costs, latency and the promised release of its weights.
Hirundo says it made Alibaba’s Qwen more politically neutral without reducing capability, but independent testing must determine whether the change holds across unfamiliar prompts and real-world evaluations.
Reflection AI’s Beam targets the open-model efficiency race, promising frontier-style reasoning and coding performance with far lower inference costs, though its benchmark and compute claims remain unverified.
Anthropic’s Claude Haiku 5.5 enters the low-cost AI race with a claimed 90% price reduction, challenging OpenAI’s GPT-6 Luna for high-volume enterprise workloads and everyday automation.
OpenAI’s GPT-6 rollout pairs a new model with Intelligent UI, testing ChatGPT’s product execution, subscription strategy and ability to turn advanced AI into everyday workflows.
TwelveLabs’ Pegasus 1.6 analyzes first-person work videos to extract actions, objects, hand movements and task boundaries, targeting robotics’ costly shortage of structured training data for physical AI.