A lot happened in AI this week. Here’s what matters and why.
OpenAI: GPT-5.5 Pro Launches
OpenAI released GPT-5.5 Pro to API users. Key improvements over GPT-5.5:
- Better reasoning: 12% improvement on complex multi-step tasks
- Reduced hallucinations: 23% fewer factual errors than GPT-5.5
- Longer context: Extended to 2M tokens for Pro users
- Pricing: $15/$60 per million tokens (input/output) — 3x the cost of GPT-5.5
Who should care: Developers building production applications where accuracy matters more than cost.
Google: Gemini 3.5 Flash Widely Available
Google made Gemini 3.5 Flash available to all users after I/O 2026. Highlights:
- Speed: 2x faster than Gemini 3.1 Flash
- Cost: $0.075/$0.30 per million tokens (cheapest in class)
- Multimodal: Now handles video input natively
- Google Workspace: Deep integration with Docs, Sheets, Gmail
Who should care: Anyone in the Google ecosystem. Gemini 3.5 Flash is now the cheapest high-quality model available.
Anthropic: Claude Gets Memory
Anthropic rolled out memory features for Claude:
- Persistent preferences: Claude remembers your writing style and preferences
- Project context: Remembers details across conversations in Projects
- Privacy controls: Full control over what Claude remembers
Who should care: Heavy Claude users who want consistent, personalized responses.
Codex CLI Goes Open Source
OpenAI open-sourced Codex CLI, their autonomous coding agent:
- Community contributions: Developers can now extend Codex with custom tools
- MCP protocol: Standardized tool integration
- Self-hosting: Run Codex on your own infrastructure
Who should care: Developers who want to customize their AI coding workflow.
Image Generation Updates
Midjourney v7.1 Patch
- Improved text rendering (still not perfect)
- Better consistency for brand guidelines
- Faster generation times (~20s average)
Ideogram 3.0 Launch
- Near-perfect text in images
- Multiple font support
- Logo generation improvements
What This Means for You
For developers: GPT-5.5 Pro is worth the premium for production apps. Codex CLI open source is a big deal.
For content creators: Gemini 3.5 Flash is the new budget king. Claude’s memory feature is a game-changer for long-form writers.
For businesses: The cost of AI is dropping fast. Gemini 3.5 Flash at $0.075/M tokens makes AI accessible to even small businesses.
Looking Ahead
- GPT-5.6: Rumored for late 2026. Codex logs briefly referenced it.
- Claude Opus 5: Expected Q3/Q4 2026
- Gemini 4: Google’s next generation, likely 2027
Stay tuned for next week’s update.