Open-weight vs closed API models
When to self-host or buy open weights versus calling a frontier API. Catalog flags and trade-offs.
What “open” means on this site
isOpenSource on CompareLLM means we marked the row as open-weights (or open-weight-adjacent). It is not a lawyer’s license review. Read the actual license before you ship.
Closed rows are hosted APIs. You cannot move the weights. You can usually move the prompts.
When closed still wins
If you need computer-use, the latest vision stack, or a vendor SLA, the closed frontier models still lead most preference Elo tables. Pay the token bill or accept a smaller model.
When open-weights win
High volume, data residency, or “we will fine-tune this” are the usual reasons. Filter the leaderboard to open weights and open the Stack Engine open-weight reasoning preset.
Ready to evaluate your stack?
Calculate your optimal model weights with Stack Engine or compare top models head-to-head.
Related guides
Same topic, next level of detail.
The Complete Guide to Local AI Hardware, VRAM Sizing & Quantization in 2026
Everything you need to know about running open-weight LLMs locally: GPU VRAM vs Apple Unified RAM, Quantization (FP16 down to Q2), KV Cache scaling, Layer Offloading, and Inference Speed (tok/s).
Reasoning Effort, Dynamic CoT, and Multi-Tier Model Derivatives in 2026
How test-time compute, reasoning effort controls, fast routes, and derivative tiers (Sol/Terra/Luna, Opus/Fable/Sonnet, DeepSeek V4, Qwen 3) affect benchmarks and production bills.
How to pick an LLM in 2026
A practical order of operations: task, budget, latency, then Elo. Links into CompareLLM stacks and compares.
Explore More AI Intelligence Tools
Pairwise Model Comparisons
Compare any two models on our category ratings, speed classes and token pricing.
Curated 'Best-Of' Indexes
Targeted rankings for Best Coding LLMs, Best Cheap APIs, Shortest Wait, and top-tier models.
AI Hardware Calculator
Interactive VRAM calculator, quantization levels (FP16, Q8, Q4), KV cache context, and local hardware fit.
Live Market Intelligence
Live audit trail of benchmark updates, new model releases, and API price cuts.
Interactive Model Finder
Answer 5 quick questions to compute deterministic model recommendations for your use case.
Methodology & Standards
How a CompareLLM rating is computed, what evidence it admits, and how prices and speed classes are recorded.