A thinking model designed for a balance between price and performance. It builds upon Gemini 2.0 Flash with upgraded reasoning, hybrid thinking control, multimodal capabilities (text, image, video, audio input), and a 1M token input context window.
发布日期2025年5月20日
参数规模—
上下文长度1.0M
许可证Proprietary
知识截止2025年1月31日
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| Global-MMLU-Lite | general language reasoning | 88.4 | 来源 |
| AIME 2024 | math reasoning | 88.0 | 来源 |
| FACTS Grounding | reasoning | 85.3 | 来源 |
| GPQA | reasoning general | 82.8 | 来源 |
| MMMU | multimodal reasoning general | 79.7 | 来源 |
| AIME 2025 | math reasoning | 72.0 | 来源 |
| Vibe-Eval | multimodal vision general | 65.4 | 来源 |
| LiveCodeBench v5 | reasoning general | 63.9 | 来源 |
| Aider-Polyglot | general code | 61.9 | 来源 |
| SWE-Bench Verified | reasoning frontend_development code | 60.4 | 来源 |
| Aider-Polyglot Edit | general code | 56.7 | 来源 |
| MRCR | long_context reasoning general | 32.0 | 来源 |
| SimpleQA | general reasoning | 26.9 | 来源 |
| Humanity's Last Exam | general | 11.0 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| QiHang | $0.09 | $0.71 | 1.0M | — | — | ✓ | ✗ | ✗ |
| Jiekou.AI | $0.27 | $2.25 | 1.0M | — | — | ✓ | ✗ | ✗ |
| 302.AI | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| Abacus | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| AIHubMix | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| Auriko | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| FrogBot | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ | |
| Vertex | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| Helicone | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| LLM Gateway | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
| NanoGPT | $0.30 | $2.50 | 1.0M | — | — | ✗ | ✗ | ✗ |
| SAP AI Core | $0.30 | $2.50 | 1.0M | — | — | ✓ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。