In testing completed on September 3, 2026, Gemini 3.6 Flash and GPT-5 Mini earned maximum scores of 100% on the seven-day standing benchmark while generating zero total spend. Both top-scoring models ran alongside a field where every tracked endpoint maintained a 100% uptime rate across hourly checks. While Gemini 3.6 Flash posted a 2.60s median response latency, GPT-5 Mini followed closely at 2.67s, positioning both models at the top of overall accuracy metrics without requiring direct query fees under current test parameters.
Speed
Latency results across the 24-hour evaluation period revealed a clear split between lightweight efficiency models and larger architectures. [Gemini 2.5 Flash Lite](/models/google-gemini-2-5-flash-lite) recorded the fastest performance overall with a 660ms median response time, followed by Gemini 3.1 Flash Lite at 764ms and GPT-5.4 Mini at 783ms. Gemini 2.5 Flash also remained below one second, registering an 844ms median response time.
Higher-tier and mid-range systems occupied the sub-two-second and three-second windows. GPT-5.6 Luna reached a 1.02s median latency, GPT-5.6 Terra logged 1.30s, and GPT-5.6 Sol registered 1.33s. Lower-tier speed results included Gemini 3.5 Flash at 2.29s, Gemini 3.6 Flash at 2.60s, GPT-5 Mini at 2.67s, and GPT-5 Nano at 2.80s. The slowest endpoints in the check cycle were Gemini 3.1 Pro (preview) at 3.50s and Gemini 2.5 Pro at 3.72s median response time. All 13 endpoints maintained 100% uptime across hourly checks throughout the full 24-hour sampling window.
Prices
Pricing structure telemetry for Google endpoints reflects wide variance depending on model tier and input-output channel allocation. Gemini 2.5 Flash Lite offered the lowest baseline input cost at $0.100/M tokens and an output rate of $0.400/M tokens. Gemini 3.1 Flash Lite registered at $0.250/M input and $1.50/M output, while Gemini 2.5 Flash was recorded at $0.300/M input and $2.50/M output.
The table below details the observed billed pricing per million tokens across measured Google models:
| Model | Input Price ($/M) | Output Price ($/M) |
|---|---|---|
| Gemini 2.5 Flash Lite | $0.100 | $0.400 |
| Gemini 3.1 Flash Lite | $0.250 | $1.50 |
| Gemini 2.5 Flash | $0.300 | $2.50 |
| Gemini 2.5 Pro | $1.25 | $10.00 |
| Gemini 3.5 Flash | $1.50 | $9.00 |
| Gemini 3.1 Pro (preview) | $2.00 | $12.00 |
Higher-tier models showed substantial cost premiums. Gemini 2.5 Pro costs $1.25/M input tokens and $10.00/M output tokens. Gemini 3.5 Flash sits at $1.50/M input and $9.00/M output. The highest observed rate belonged to Gemini 3.1 Pro (preview), priced at $2.00/M input tokens and $12.00/M output tokens.
Benchmark
Over the seven-day standing evaluation period, accuracy totals demonstrated strong performance among lightweight variants. Behind the 100% standing scores achieved by Gemini 3.6 Flash and GPT-5 Mini ($0 total spend), GPT-5.6 Terra secured 89.3% of available points at $0 spend. Gemini 3.1 Flash Lite recorded an 86.5% benchmark result at $0 spend, outperforming Gemini 2.5 Flash at 79.2% ($0 spend) and GPT-5.6 Luna at 71.9% ($0 spend).
These telemetry readings demonstrate explicit operational output, latency, and cost parameters recorded over designated sampling windows. They do not predict future provider price changes, long-term system stability under unmonitored load spikes, or domain-specific task accuracy outside the configured benchmark suite.
