AI News

IBM announced one new Granite Speech model and revisions to two others, targeting multilingual translation, faster transcription and speaker labeling.

A report says SenseTime open-sourced an embodied AI harness and a 27B robot model, but release access, licensing and capabilities remain unverified.

A BigGo Finance headline pairs Musk’s superintelligence warning with a claim about Grok on the iPhone side button; details are unavailable.

TestingCatalog reports a Claude Fable 5.5 release next week, but Anthropic has not confirmed the model or timing in the available information.

Apertus 1.5 adds image and experimental audio input, a 262,144-token context window and updated training to its open-weight 8B and 70B models.

Meta’s open-weight multimodal model targets agentic coding, with a one-million-token context and infrastructure demands that require substantial memory.

Anthropic co-founder Jack Clark told an IMF audience that AI may accelerate science and urged policymakers to prepare for rapid change.

MiniMax launched M3.1 Flash at lower token prices, but independent measures and a 24-article newsroom test show trade-offs.

A headline claims Claude Opus 5.5 beat a human baseline and shifted AI self-improvement forecasts, but provides no supporting details.

NVIDIA’s open-weight Nemotron 3.5 Bolt scores near its 550B flagship on one benchmark, but ranks fourth under 100B and uses 82,000 tokens per task.