Model Skills

Street smart. Best for real-world agentic tasks. Highest writing quality of any frontier model.

OSWorld computer use leader (75.3% vs 72.4% human baseline). Strong vision. Tool use native.

Best-in-class context window. Excellent at processing huge documents. Google's current enterprise flagship.

Anthropic's most capable. Overkill for most tasks but unbeatable on hard reasoning and nuanced writing.

Benchmark workhorse. Fast, capable, strong vision. The default for most OpenAI integrations.

Thinking mode enabled. Extended reasoning for hard problems. Most expensive OpenAI model.

Zhipu AI's flagship. Competitive across all dimensions. Strongest on Chinese-language reasoning and structured tasks.

Truth-seeking first. Strong on factual accuracy, real-time X data access, and hard reasoning tasks.

Good balance of writing quality and long context. Strong on creative and structured output tasks.

Strong on very long contexts and Chinese-language tasks. Competitive mid-tier pricing.

1M token context window. Designed for massive document processing pipelines.

Fast mid-tier option with strong Chinese language support and solid long-context handling.

Fast and cheap. Ideal for high-volume, low-complexity tasks where cost matters more than quality.

The model that broke Twitter in Jan 2026. Matches o1 on reasoning at a fraction of the cost. No vision.

EU-based, GDPR-native. Strong for enterprises that can't send data to US providers.

Best open-weight model for code. Rivals frontier closed models at zero API cost when self-hosted.

Best value coding model. Near-frontier coding at mid-tier cost. No vision capability.

Best open-weight model for self-hosted deployments. Run locally, zero data exposure, no API costs.