Gemini 2.5 Pro
Google's proven reasoning model for coding, math, and multimodal analysis
Shogo Agent Index
Supported#81 of 9224.5range 19.4–29.6 · 53% of weight measured
Agentic coding
27
57% measured
Long-horizon autonomy
32
80% measured
Knowledge-work agents
21
68% measured
Computer use
13
estimated
Security & context learning
24
estimated
Price
- Input
- $1.25/M
- Output
- $10.00/M
- Cache read
- $0.13/M
- 3:1 blend
- $3.44/M
- Agent-effective
- $0.76/M
Cheapest host for agents: Google AI Studio at $0.38/M.
Agent-effective = 20 input tokens per output token, 85% cache hits. Why.
Agentic benchmarks
| Benchmark | Score | Normalised | Index weight | Harness / source |
|---|---|---|---|---|
| SWE-bench Verified | 57.6% | 8% | source | |
| Terminal-Bench 2.0 | 32.6% | 8% | Terminus 2 | |
| DeepSWE | est. | 6% | predicted 16 | |
| FrontierSWE | est. | 5% | predicted 2 | |
| CursorBench | est. | 4% | predicted 22 | |
| GSO-Bench | 3.9% | 4% | OpenHands | |
| METR Time Horizons | 55.4% | 10% | source | |
| Vending-Bench 2 | $574 | 6% | source | |
| PostTrainBench | est. | 4% | predicted 14 | |
| APEX-Agents | est. | 8% | predicted 17 | |
| GDPval | 23.3% | 7% | source | |
| Remote Labor Index | 0.8% | 5% | source | |
| DeepResearch Bench | 42.8% | 5% | source | |
| OSWorld 2.0 | est. | 6% | predicted 1 | |
| OSWorld | est. | 4% | predicted 29 | |
| Cybench | est. | 5% | predicted 33 | |
| CL-bench | est. | 5% | predicted 15 |
Other benchmarks
| Benchmark | Category | Score | Source |
|---|---|---|---|
| Aider Polyglot | Coding | 83.1% | source |
| AlgoTune | Coding | 1.51× | source |
| ARC-AGI | Reasoning | 41.0% | source |
| ARC-AGI-2 | Reasoning | 4.9% | source |
| BALROG | Long-horizon | 43.3% | source |
| CritPt | Reasoning | 2.0% | source |
| DTBench | Knowledge | 82.4% | source |
| Fiction.LiveBench | Long context | 91.7% | source |
| FrontierMath Tiers 1-3 | Math | 24.6% | source |
| FrontierMath Tier 4 | Math | 0.0% | source |
| GeoBench | Multimodal | 86.0% | source |
| GPQA Diamond | Reasoning | 85.3% | source |
| Humanity's Last Exam | Reasoning | 21.6% | source |
| LMCA | Knowledge | 34.8% | source |
| MATH Level 5 | Math | 95.9% | source |
| OTIS Mock AIME | Math | 84.7% | source |
| SciCode | Coding | 46.3% | source |
| SimpleBench | Reasoning | 62.4% | source |
| TheAgentCompany | Knowledge work | 30.3% | source |
| VPCT | Multimodal | 48.0% | source |
| WebDev Arena | Coding | 1226 | source |
API hosts
Live prices and uptime for every host serving Gemini 2.5 Pro via OpenRouter.
| Host | Input | Output | Cache read | Agent $/M | Uptime 24h | Quant |
|---|---|---|---|---|---|---|
| Google AI Studio | $0.63 | $5.00 | $0.063 | $0.38 | 100.0% | — |
| Google AI Studio | $1.25 | $10.00 | $0.13 | $0.76 | 100.0% | — |
| Google AI Studio | $2.25 | $18.00 | $0.23 | $1.36 | 100.0% | — |