Microsoft’s Phi-4: Enterprise-Grade AI, Now On-Device
Microsoft’s Phi-4 small language models deliver enterprise-grade reasoning directly on local devices—no cloud dependency required. This advancement improves data security, accelerates AI adoption in controlled environments, and enhances scalability by reducing reliance on centralized services.
Key features:
- Outperforms DeepSeek-R1-70B and OpenAI o1-mini on AIME 2025, a PhD-level STEM benchmark
- Phi-4-reasoning-plus: Advanced training techniques including reinforcement learning for elevated reasoning performance
- Phi-mini: Optimised for edge devices and low-latency on-device use
- Already in production: integrated within Windows Copilot+ PCs to enable offline summarisation and application-level intelligence
Canva Visual Suite 2.0: AI-Powered All-in-One Workspace
Canva’s all-in-one AI workspace combines content creation, data visualisation, and no-code app building. This streamlined environment serves marketing, content and operations teams by reducing tool-switching and accelerating output.
Key features:
- AI assistant generates content, visuals, presentations and answers file-specific questions
- Canva Sheets + Magic Studio: transforms data into animated, branded visuals
- Canva Code: build calculators, tools or games without writing code
- Unified workflow for design, documentation, slides and app development
- Single interface from brainstorm to output, aligning teams and speeding time to impact
Check out our blog on — What Senior Leaders Must Know About Recent AI Developments
OpenAI Codex: Secure AI Agent for Software Development
OpenAI’s Codex is a cloud-based AI agent optimised for software development, automating complex and repetitive coding tasks so development teams can focus on higher-value activities while maintaining security and control.
Key features:
- Operates without internet access during tasks to protect code integrity
- Transparent workflow with logs and test results
- Light-weight Codex-mini variant for local development at lower cost
- Seamless integration into existing developer pipelines
- Used by leading organisations for faster, more secure development cycles
Windsurf SWE-1: AI That Understands Engineers
An engineering-domain-specific AI model built for end-to-end software engineering workflows within the Windsurf IDE. Designed to deliver faster code delivery, higher code acceptance rates and lower cost outcomes for engineering teams.
Key features:
- Trained on real-world developer usage data
- Matches leading open models on SWE benchmarks
- Out-performs peers like DeepSeek and Qwen in domain-specific metrics
- Improves metrics such as Daily Lines Contributed per User and Developer Cascade Contribution Rate
- Continuously raising developer productivity and lowering iteration time
Teams adopting new AI workspaces often need structured AI Consultation to identify the right use cases, integrate tools, and maximise productivity gains.
Each month, we curate executive briefings on the latest in AI and analytics for senior leaders in our CXO Analytics Newsletter.
At Perceptive Analytics, we partner with enterprises to turn data into strategic advantage through Advanced Analytics Consulting, Generative AI Consulting, and Business Intelligence Solutions. To explore how AI could drive measurable impact in your organisation, book a free consultation.