Google bumped Gemini 2.5 Flash to a 2M-token context, matching Pro and leaping past Claude Opus 4.5's 1M.
Bundled with the release: Grounding API now supports multi-step retrieval and citation deduplication across reasoning chains.
Gemini 2.5 Flash Hits 2M Token Context
Google expands context parity with Claude and adds Grounding API improvements in the same drop.
FAQ
How does Gemini 2.5 Flash now compare to Claude Opus 4.5 on context length?
Gemini 2.5 Flash sits at 2M tokens, double Claude Opus 4.5's 1M ceiling. Pro also runs at 2M, so Flash now achieves parity with the larger Gemini tier on context window alone.
What changed in the Grounding API?
Two upgrades: multi-step retrieval (the model can pull additional sources mid-chain) and citation deduplication so reasoning chains do not repeat the same source under multiple identifiers.
Is Gemini 2.5 Flash priced differently after the context upgrade?
Google did not change Flash's standard rate in the same drop. The 2M context is available to all Flash users at existing pricing tiers.