Nano Banana 2.1 Is Out: Google’s Best Image Model Yet Cuts API Costs In Half

Google on Tuesday launched Nano Banana 2.1, the most recent model of its picture era and enhancing mannequin, delivering broad enhancements in visible high quality, immediate adherence, textual content rendering and multi-turn character consistency whereas sustaining Flash-level velocity and value effectivity. The mannequin serves because the extra environment friendly counterpart to Gemini 3 Pro Image, often known as Nano Banana Pro.

The replace is rolling out throughout the Gemini app, Google Search’s AI Mode, Google Ads and developer instruments together with Google AI Studio, Flow and Stitch. It arrives at a time when the Nano Banana household has develop into a central pillar of Google’s shopper AI technique: in September 2025, the unique mannequin’s potential to show selfies into collectible collectible figurines propelled Gemini to the highest of each main app shops, ending ChatGPT’s practically three-year dominance and coinciding with Alphabet’s market worth surpassing $3 trillion. Version 2 adopted in February, constructed on Gemini 3.1 Flash Image, and launched grounding in Google Search in order that photos of actual occasions or folks can be extra correct.

Sharper Editing and Stronger Benchmarks

Google highlights three headline upgrades within the new launch: improved visible design, extra exact mask-based enhancing that adjustments solely a marked area of a picture, and stronger topic consistency so an individual or object stays recognizable throughout a number of edits. On paper, the mannequin handles as much as 14 reference photos concurrently, sustaining consistency for as much as 4 characters and constancy for as much as ten objects. It outputs at 1K, 2K and 4K resolutions, fixes tiling artifacts on excessive panoramic facet ratios reminiscent of 1:4, 1:8, 4:1 and eight:1, and renders textual content and infographic layouts extra precisely.

The positive aspects are seen in benchmarking as nicely. On total choice in text-to-image checks, Nano Banana 2.1 scored 1,050 ELO factors, in contrast with 990 for Nano Banana 2 and 935 for Nano Banana Pro.

Developers acquire further management by means of configurable pondering ranges, starting from minimal to high, and thru grounding with Google Web and Image Search, successfully permitting the mannequin to search for info earlier than producing a picture.

The effectivity positive aspects prolong to price. Through Google’s developer API, a typical 1K picture prices $0.0336 — roughly half the $0.067 charged for Nano Banana 2 — whereas a 4K picture runs $0.0756 versus $0.151 beforehand. Batch processing receives an extra 50% low cost. In sensible phrases, producing a thousand standard-resolution photos now prices about $33.60, down from $67, positioning Nano Banana 2.1 as each a high quality chief and one of the cost-effective choices within the generative picture market.

The publish Nano Banana 2.1 Is Out: Google’s Best Image Model Yet Cuts API Costs In Half appeared first on Metaverse Post.