Google's Gemini 3.5 Flash costs 5.5 times more to run than its predecessor, following price increases across the AI industry.

Google released Gemini 3.5 Flash with token prices tripled: $1.50 per million input tokens and $9.00 per million output tokens, up from $0.50 and $3.00 for Gemini 3 Flash. Despite lower per-token pricing than Gemini 3.1 Pro, the model consumes so many more tokens on agent-based tasks (tasks where the AI plans and uses tools) that total costs end up 75 percent higher than Gemini 3.1 Pro.

On agent tasks, Gemini 3.5 Flash achieved an Elo score of 1,656 on GDPval-AA, a massive leap over Gemini 3 Flash (1,204), but it requires an average of 49 turns per task—more than any other model tested—driving input token consumption sharply up. It also scores 55 on the Artificial Analysis Intelligence Index, nine points above Gemini 3 Flash, and its hallucination rate drops to 61 percent from 91 percent, though it trails top models at 25 percent.

Coding remains a weakness: Gemini 3.5 Flash scores just 45 on the Artificial Analysis Coding Index, well behind Gemini 3.1 Pro Preview (55) and far behind GPT-5.5 (59) and GPT-5.4 (57). The model supports video and audio input alongside text and images, and clocks over 280 output tokens per second, roughly 70 percent faster than Gemini 3 Flash.

Anthropic's Opus 4.7 saw a hidden price increase of roughly 30 to 40 percent over its predecessor due to higher token consumption, and OpenAI's GPT 5.5 jumped about 50 to 90 percent over 5.4, reflecting a broader industry trend toward pricier models built for multi-step agentic tasks.