Google Launches Gemini 3.1 Flash-Lite at 1/8th Pro Cost

🔑 Enhanced Key Takeaways

• •Gemini 3.1 Flash-Lite supports multimodal inputs including text, up to 3,000 images (7MB inline), 10 videos (up to 1 hour), and 8.4 hours of audio, with a 1M input token context window.[1][2][4]
• •Pricing is $0.25 per 1M input tokens and $1.50 per 1M output tokens, optimized for high-volume agentic tasks and low-latency applications.[3][4]
• •Achieves 60.1% on MRCR v2 (8-needle) long context benchmark at 128k tokens and 21.0% at 1M tokens, outperforming several peers in needle retrieval.[2]
• •Features include function calling, structured outputs, thinking, caching, batch API, code execution, and search grounding, but lacks audio/image generation and computer use.[4]

🛠️ Technical Deep Dive

• •Maximum input tokens: 1,048,576; maximum output tokens: 65,535 (or 65,536 in preview).[1][2][4]
• •Natively multimodal: supports text, images (up to 3,000 per prompt, 7MB inline/30MB GCS), documents (3,000 files, 1,000 pages/file, 50MB), video (10 files, ~1hr no audio), audio (8.4hrs or 1M tokens).[1][2][4]
• •Architecture based on Gemini 3.1 Pro, inheriting its limitations and acceptable usage policies; suited for high-volume, low-latency tasks but less capable than Pro.[2]
• •Supported MIME types include image/png/jpeg/webp/heic/heif, video/mp4/webm/etc., audio/mp3/wav/etc., and documents pdf/text/plain.[1]
• •Capabilities: function calling, structured outputs, thinking levels, file search, URL context, search grounding; no image/audio generation, live API, or computer use.[4]

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline