LLM Leaderboard

Most LLM leaderboards show a single metric. Quality score, or ELO rating, or some proprietary number. That tells you which model is smartest but nothing about whether you can afford to use it.

This leaderboard combines four dimensions: quality (from the Artificial Analysis Intelligence Index), price per million tokens (from OpenRouter and official APIs), generation speed (tokens per second), and context window size. The Value Score combines quality and price into a single number so you can find models that punch above their cost.

We exclude deprecated models, extremely limited-availability models, and models that don't offer meaningful advantages over alternatives in their price range. The goal is a useful ranking, not an exhaustive list.

The top of the leaderboard is tighter than ever. Claude Fable 5 edges out GPT-5.6 Sol by a single point (60 vs 59), but costs nearly twice as much and uses 2.8x more tokens per task. For most workloads, Sol delivers equivalent results at lower total cost.

The real story is the mid-tier. Grok 4.5 (score 54) and Kimi K3 (score 57) sit between the frontier models and budget options. They handle 90% of tasks that premium models can, at 20-50% of the price. If you're running agents or batch processing, these are where the economics make sense.

Budget models have gotten good enough for production. DeepSeek V4 Flash at $0.42/1M total tokens handles simple classification, summarization, and formatting tasks that used to require GPT-4 class models. GPT-5.6 Luna at $7/1M total is the new default for "good enough" work.

For deeper analysis of each model, see our Best LLM in 2026 guide, coding model rankings, GPT-5.6 breakdown, or cost comparison. Use the Model Selector for personalized recommendations.