Microsoft has launched Microsoft-Decision-1, a 9B parameter model derived from Qwen3.5-9B, designed for fixed-choice decision-making rather than text generation. This model, available in Foundry, aims to provide calibrated probabilities for options, potentially streamlining agent workflows by replacing probabilistic text generation with a more structured classification call. Separately, Anthropic's Claude Code has seen updates to its guard hooks, with version 2.1.295 introducing a 'block' option for onFailure, ensuring that failed hooks prevent actions rather than allowing them to proceed. This change enhances the reliability of policy enforcement within the system. Additionally, Anthropic's Haiku 5.5 model is now listed with a 1M-token context window and significantly reduced pricing, though a 100K token limit for prompts and a new tokenizer that increases token counts for the same text may affect the actual cost savings. AI
IMPACT New decision-making models could streamline agent workflows, while safety improvements in Claude Code and clarified pricing for Haiku 5.5 impact developer cost and trust.
RANK_REASON Multiple product updates and model releases from significant AI players. [lever_c_demoted from significant: ic=1 ai=1.0]
- Anthropic
- Claude Code
- Haiku 4.5
- Haiku 5.5
- LiteLLM
- Microsoft
- Microsoft-Decision-1
- Qwen3.5-9B
- SGLang
- Unsloth
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
