01Coding model efficiency
Microsoft launches MAI-Code-1.1-Flash for faster, lower-cost coding assistance
Microsoft says MAI-Code-1.1-Flash delivers higher code quality with 25% greater token efficiency and one-quarter the cost of the model it launched in June
Microsoft says MAI-Code-1.1-Flash delivers higher code quality with 25% greater token efficiency and one-quarter the cost of the model it launched in June
Coding-model economics are becoming part of the developer-tool product, not just a backend benchmark
- Watch next
- Watch rollout notes and independent task results before treating the efficiency claim as a production baseline.
02Open agent model infrastructure
NVIDIA publishes Nemotron 3.5 Lightning for long-running agents
NVIDIA published Nemotron 3.5 Lightning for a long-running agent workflow, with open checkpoints and a reported 4x output-speed benchmark
NVIDIA published Nemotron 3.5 Lightning for a long-running agent workflow, with open checkpoints and a reported 4x output-speed benchmark
A smaller active-parameter path may change the cost and latency tradeoff for always-on agent services
- Watch next
- Watch checkpoint licenses, reproducible task results, and hardware-specific latency before deployment decisions.
03Enterprise AI governance
Anthropic extends Compliance API coverage to Claude Cowork and Claude Code
Anthropic added Compliance API coverage for Cowork desktop, Cowork web, and Claude Code, exposing programmatic audit access for enterprise teams
Anthropic added Compliance API coverage for Cowork desktop, Cowork web, and Claude Code, exposing programmatic audit access for enterprise teams
Programmatic audit coverage can reduce the gap between AI-tool adoption and enterprise governance requirements
- Watch next
- Watch API scope, retention rules, and which activity types are actually available to each enterprise plan.