My current stack, favorite models, and the tools I actually use every day. Updated as things change, which is constantly.
Last updated: June 2026
01Anthropic's latest and strongest frontier model, and the Anthropic model I'm actively using right now. Best when I want Claude-style taste, speed, and agentic coding instincts with the newest reasoning ceiling.
Latest Anthropic frontier model
Strong for agentic coding and product work
Excellent taste for UI and writing
High ceiling on hard reasoning tasks
New enough that pricing and behavior may shift
Worth reserving for complex or high-leverage work
02My default OpenAI model. Best at long-running agentic coding tasks. Slower and more deliberate. Thinks before it acts, takes time to understand the codebase. Doesn't default to action like Opus. The model you want for complex backend work.
Best for complex, multi-step coding tasks
Deliberate, thinks before acting
Strong at understanding large codebases
Slower than Opus
UI output not as polished
03Still one of the best at UI and front-end work. Nicest to talk to: natural, fluid, pleasant. 1M token context window. Fast, prone to action. Very expensive on API, incredible value on a subscription.
Best at UI and frontend code
1M token context window
Natural, fluid conversation style
Fast and action-oriented
Very expensive on API
Can be too eager to act before thinking
04Very solid model, right on the Pareto frontier for intelligence and cost. State of the art at tool calling, with a 1M token context window and extremely aggressive API pricing.
State-of-the-art tool calling
Excellent intelligence per dollar
1M token context window
Very low API pricing
Not frontier for coding
xAI ecosystem is less mature than Anthropic and OpenAI
05The multimodal leader. Best for video, images, and massive documents. Useful when you need to process visual content or very large files.
Best multimodal capabilities (video, image, audio)
Excellent at structured output from messy input
Huge context window
Unreliable at tool calling in agent loops
Less predictable in complex harnesses
0690% of Opus intelligence at a fraction of the cost. Open source. The sleeper pick if you need near-frontier reasoning without frontier pricing.
Near-frontier intelligence, fraction of the cost
Open source
Great value for high-volume workloads
Not quite frontier on the hardest tasks
Smaller ecosystem and tooling