ModelBench

Leads SWE-bench Verified and Text Arena coding. Best for software engineering.

Anthropic's top model. Excels at agentic terminal tasks and web development.

OpenAI's latest flagship with extended thinking. Top-tier on GPQA and competitive on SWE-bench.

Fast, cheap model that punches above its weight on coding benchmarks.

Maximum compute reasoning variant. Leads FrontierMath tiers and OTIS Mock AIME.

Leads SimpleBench and FrontierMath Tier 4. Anthropic's reasoning specialist.

Google's flagship with industry-leading context and Humanity's Last Exam score.

Strong all-rounder at a lower price point than 5.5 series.

Specialized coding model tuned for IDE and agentic development workflows.

Best price-performance ratio in the Claude family.

Top open-weight contender with API access at aggressive pricing.

Reasoning-focused model with strong math and science performance.

Strong coding arena performer with open-weight availability.

xAI flagship with strong long-context comprehension and real-time X integration.

Optimized for latency-sensitive applications with live data access.

Previous-gen workhorse still widely deployed for production workloads.

European flagship with strong multilingual capabilities.

Pioneering open reasoning model. Strong math, weaker on agentic tasks.

Meta's open multimodal model for on-premise and fine-tuning workflows.

Enterprise-focused model optimized for retrieval-augmented generation.