Context window sizes have expanded dramatically. Among 100 coding models, the largest context window is 2.0M tokens, the median is 262.1K tokens, and 100 models support 128K+ tokens. 32 models support over 1M tokens.
Largest Context Windows
Does More Context = Better Performance?
A larger context window enables processing entire codebases, long documents, and multi-turn conversations. However, raw context size does not directly correlate with overall model quality. Many top-scoring models have moderate context windows but excel in capability depth and output quality.
The real differentiator is not just the size of the context window but how well the model retrieves and reasons over information within it. Needle-in-a-haystack benchmarks show that many models degrade significantly at the edges of their context window, even when they technically support the input length.