Introducing Muse Spark 1.1

24 days ago

Meta launches Muse Spark 1.1, an upgraded multimodal reasoning model for agentic tasks with improved tool/computer use, coding, and multimodal understanding.
The model features a 1-million token context window, multi-agent orchestration for faster project handling, and public preview via the Meta Model API.
Muse Spark 1.1 excels in computer-use workflows, coding (debugging, migrations), multimodal perception, and safety, with strong adversarial robustness.
Early partners like Replit and Box praise its comprehensive capabilities for large-scale agentic workloads and enterprise applications.
Meta highlights ongoing research momentum, with more advanced models in training, advancing toward personal superintelligence.

Laguna S 2.1 is a 118B-parameter Mixture-of-Experts model with 8B active parameters per token and a 1M-token context window, developed in under nine weeks.
It achieves competitive scores against much larger models on long-horizon coding benchmarks, scoring 70.2% on Terminal-Bench 2.1 and 40.4% on DeepSWE v1.1.
All evaluation trajectories are publicly released at trajectories.poolside.ai for transparency and reproducibility.

Introduction of Kimi K3: a 2.8T-parameter open model with native vision and a 1-million-token context window.
Architecture based on Kimi Delta Attention and Attention Residuals, with efficient scaling and 16 active experts out of 896.
Strong performance in long-horizon coding, including kernel optimization, compiler development, and game dev with visual reasoning.

vLLM Semantic Router enables collaborative micro-agents within a single model API call, enhancing AI inference by orchestrating multiple models.
The router acts as an active serving layer, selecting optimal collaboration patterns like Confidence, Ratings, ReMoM, Fusion, or Workflows based on request analysis.
Key benefits include cost savings, safety enforcement, and improved performance without exposing complexity to users, maintaining a single API surface.

Grok 4.525 days ago

Grok 4.5 is released in collaboration with SpaceXAI as the most intelligent model, designed for tasks beyond software engineering.
It can handle difficult, long-running tasks in various fields like software engineering, data science, finance, and legal work by creatively using tools.
Cursor subscription plans include significant model usage with double usage for the first week and added safeguards for cybersecurity.

MAI-Cyber-1-Flash is a compact, code-heavy security model designed to find challenging vulnerabilities in complex codebases.
It is integrated into MDASH, a multi-agent vulnerability identification and remediation harness, delivering world-class performance at 50% of the cost of leading models.
The system handles up to 90% of tasks efficiently, reserving larger models for the hardest 10%, achieving a 96% score on CyberGym (12 points above Mythos).

Laguna S 2.1 is a 118B-parameter Mixture-of-Experts model with 8B active parameters per token and a 1M-token context window, developed in under nine weeks.
It achieves competitive scores against much larger models on long-horizon coding benchmarks, scoring 70.2% on Terminal-Bench 2.1 and 40.4% on DeepSWE v1.1.
All evaluation trajectories are publicly released at trajectories.poolside.ai for transparency and reproducibility.

Introduction of Kimi K3: a 2.8T-parameter open model with native vision and a 1-million-token context window.
Architecture based on Kimi Delta Attention and Attention Residuals, with efficient scaling and 16 active experts out of 896.
Strong performance in long-horizon coding, including kernel optimization, compiler development, and game dev with visual reasoning.

vLLM Semantic Router enables collaborative micro-agents within a single model API call, enhancing AI inference by orchestrating multiple models.
The router acts as an active serving layer, selecting optimal collaboration patterns like Confidence, Ratings, ReMoM, Fusion, or Workflows based on request analysis.
Key benefits include cost savings, safety enforcement, and improved performance without exposing complexity to users, maintaining a single API surface.

Grok 4.525 days ago

Grok 4.5 is released in collaboration with SpaceXAI as the most intelligent model, designed for tasks beyond software engineering.
It can handle difficult, long-running tasks in various fields like software engineering, data science, finance, and legal work by creatively using tools.
Cursor subscription plans include significant model usage with double usage for the first week and added safeguards for cybersecurity.

MAI-Cyber-1-Flash is a compact, code-heavy security model designed to find challenging vulnerabilities in complex codebases.
It is integrated into MDASH, a multi-agent vulnerability identification and remediation harness, delivering world-class performance at 50% of the cost of leading models.
The system handles up to 90% of tasks efficiently, reserving larger models for the hardest 10%, achieving a 96% score on CyberGym (12 points above Mythos).

Hasty Briefsbeta