In automated measurements collected on September 4, 2026, Gemini 3.6 Flash and GPT-5 Mini achieved perfect 100% scores on the seven-day standing benchmark without incurring execution costs. Meanwhile, latency tests across 13 distinct endpoints showed [Gemini 2.5 Flash Lite](/models/google-gemini-2-5-flash-lite) delivering the fastest response times, registering a median speed of 558ms across hourly probes with a 100% success rate. Every model tracked across all providers sustained perfect uptime over the 24-hour monitoring window.
Benchmark
The seven-day standing benchmark results demonstrate a clear performance division among evaluated models. Both Gemini 3.6 Flash and GPT-5 Mini secured 100% of available points at $0 total spend. GPT-5.6 Terra ranked third overall with an 89.8% point capture rate, followed by Gemini 3.1 Flash Lite at 85.7%.
The remaining models in the benchmark evaluations recorded sub-85% performance metrics. Gemini 2.5 Flash captured 80.2% of available points, while GPT-5.6 Luna finished lowest among tracked models with a score of 74%. All six tracked models completed the standing benchmark suite with $0 in recorded spend.
| Model | Benchmark Score | Total Spend |
|---|---|---|
| Gemini 3.6 Flash | 100% | $0 |
| GPT-5 Mini | 100% | $0 |
| GPT-5.6 Terra | 89.8% | $0 |
| Gemini 3.1 Flash Lite | 85.7% | $0 |
| Gemini 2.5 Flash | 80.2% | $0 |
| GPT-5.6 Luna | 74% | $0 |
Speed
Response times measured across hourly checks in the last 24 hours revealed significant speed variations across model families and tiers. Lightweight models dominated the fastest response categories. Gemini 2.5 Flash Lite was the quickest endpoint tested, coming in at a median latency of 558ms. GPT-5.4 Mini followed as the second-fastest model at 737ms, while Gemini 3.1 Flash Lite and Gemini 2.5 Flash clustered closely behind at 860ms and 866ms respectively.
Mid-tier models settled between one and two seconds for response generation. GPT-5.6 Luna reached a 1.16s median latency, GPT-5.6 Terra recorded 1.21s, and GPT-5.6 Sol registered 1.45s. Higher-capacity models logged substantially longer latencies during testing. Gemini 3.5 Flash took 2.35s, Gemini 3.6 Flash logged 2.48s, GPT-5 Mini reached 2.62s, and GPT-5 Nano hit 2.77s. The slowest measured endpoints were Gemini 3.1 Pro (preview) at 3.63s and Gemini 2.5 Pro at 3.71s. Every single endpoint achieved a 100% check success rate across all hourly tests.
Prices
Billed costs per million tokens across Google's Gemini lineup show wide pricing spread depending on architecture tier. Gemini 3.1 Pro (preview) represents the highest input and output cost at $2.00/M input and $12.00/M output tokens. Gemini 3.5 Flash is priced at $1.50/M input and $9.00/M output tokens, while Gemini 2.5 Pro charges $1.25/M input and $10.00/M output tokens.
At lower operational tiers, Gemini 3.1 Flash Lite costs $0.250/M input and $1.50/M output. Gemini 2.5 Flash sits slightly higher on output pricing at $0.300/M input and $2.50/M output. Gemini 2.5 Flash Lite remains the lowest-cost model listed, priced at $0.100/M input and $0.400/M output tokens.
These empirical measurements prove direct performance, pricing, and latency parameters for specific API endpoints during the designated sampling window. They do not prove long-term model stability, performance under non-standard prompt workloads, or real-world user application speeds under varying network conditions.
