Google: Gemini 2.5 Pro
google/gemini-2.5-pro
Cheapest provider
$1.25 / 1M
Fastest provider (p95)
131 tok/s
Google AI Studio
Intelligence (composite)
84.6
MMLU-Pro · HumanEval · math · GPQA
Per-provider performance
Latency / throughput / uptime / price measured across providers over the last 30 minutes of live traffic. This is what proves “sourced cheapest” — Atlas mode draws on these per call to serve the cheapest path that holds quality.
| Provider | Quant | Input $/1M | Output $/1M | Latency p50 / p95 | Throughput p50 / p95 | Uptime 30m | Success |
|---|---|---|---|---|---|---|---|
| full· unknown | $1.2500 | $10.0000 | 5047ms / 28557ms | 101 / 115.9 tok/s | 98.90% | 100.0% | |
| full· unknown | $1.2500 | $10.0000 | 5047ms / 28557ms | 101 / 115.9 tok/s | 74.83% | 100.0% | |
| full· unknown | $1.2500 | $10.0000 | 5047ms / 28557ms | 101 / 115.9 tok/s | — | 100.0% | |
| Google AI Studio | full· unknown | $1.2500 | $10.0000 | 2771ms / 6870ms | 101 / 131.3 tok/s | 95.28% | 95.1% |
“—” means live telemetry hasn’t accumulated enough recent traffic for that endpoint. “undisclosed” means the provider serves the model but doesn’t expose the quantization label (typically running fp8 / int8 internally).
Intelligence breakdown
Composite score is a weighted average of public benchmarks (30% MMLU-Pro, 25% code pass@1, 25% math, 20% GPQA). Numbers come from model cards and the Artificial Analysis intelligence harness; missing components are renormalised over what’s present.
MMLU-Pro
86.0
broad reasoning
Code
80.0
pass@1 (HumanEval / LiveCodeBench)
AIME 2025
88.0
math accuracy
GPQA Diamond
84.0
hard reasoning
Source: Artificial Analysis (AA-aligned) — Gemini 2.5 Pro (MMLU-Pro 86, GPQA-D 84, LiveCodeBench 80, AIME25 88); github.com/MiniMax-AI/MiniMax-M2
How Atlas mode sources Google: Gemini 2.5 Pro
- Strict mode — pin Google: Gemini 2.5 Pro exactly and we pass it straight through, sourced from the cheapest provider above. The same model, no substitutions — currently Google at $1.25/1M.
- Atlas mode — the default. Each call is auto-optimized for the cheapest path that holds quality, at least 5% off going direct from call one and climbing as it ramps. You always see which model served the call and exactly what you saved — thumbs-down anything you don’t like for a full refund.