Six OpenAI models returned to service over the past 24 hours following failed health checks, even as uptime across every monitored endpoint locked in at 83.3%. Measurements taken on August 12, 2026, show successful responses resuming for GPT-5 Nano, GPT-5 Mini, GPT-5.4 Mini, GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna. Despite these recoveries, the uniform 83.3% success rate across both OpenAI and Google endpoints indicates widespread check failures across the testing window.
At the top of the seven-day standing benchmark, Gemini 3.6 Flash and GPT-5 Mini achieved perfect 100% scores with zero tracked execution spend.
Prices
Pricing observations reflect published input and output billed costs per million tokens across the Google Gemini model lineup.
| Model | Input Cost (/M) | Output Cost (/M) |
|---|---|---|
| [Gemini 2.5 Flash Lite](/models/google-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 |
Gemini 2.5 Flash Lite remains the lowest-cost model measured at $0.100 per million input tokens and $0.400 per million output tokens. Gemini 3.1 Pro (preview) sits at the top of Google's pricing tier at $2.00 per million input tokens and $12.00 per million output tokens.
Speed
Latency measurements taken across 24 hours of hourly checks show a wide distribution in median response times, even as reliability metrics held identical at 83.3% success across all thirteen measured endpoints.
Gemini 2.5 Flash Lite delivered the fastest median response time at 605ms, followed closely by GPT-5.4 Mini at 834ms and Gemini 2.5 Flash at 842ms. GPT-5.6 Luna recorded an 857ms median, while Gemini 3.1 Flash Lite registered at 911ms.
Slower response times were concentrated among larger and newer flagship endpoints. GPT-5.6 Sol and GPT-5.6 Terra logged median speeds of 1.03s and 1.14s, respectively. Gemini 3.5 Flash recorded a 2.25s median, paired closely with Gemini 3.6 Flash at 2.31s. GPT-5 Nano hit 2.79s, while GPT-5 Mini registered 3.21s. The slowest endpoints were Gemini 3.1 Pro (preview) at 3.80s and Gemini 2.5 Pro at 3.81s.
Benchmark
Seven-day benchmark evaluations measure accuracy against total available points alongside cumulative execution spend.
Gemini 3.6 Flash and GPT-5 Mini posted flawless 100% benchmark completion rates without incurring billed execution charges. GPT-5.6 Terra earned 89.7% of available points at $0 spent, followed by GPT-5.6 Luna at 84.8% and $0 spent. Gemini 3.1 Flash Lite captured 82.2% of points with $0.00554 in total spend, while Gemini 2.5 Flash reached 80.2% with $0.00912 spent.
These measurements reflect direct observational data from automated hourly testing and public pricing tables. They prove precise endpoint latency, price tiers, and benchmark accuracy during the specified window, but do not explain network conditions, vendor backend routing, or unobserved infrastructure changes.
