MODEL HISTORY
Gemma 4 E4B (Reasoning) 指标历史
Gemma 4 E4B (Reasoning) 的公开评测事实历史,共 65 条观测、37 项指标;只记录来源变化,不插值、不合成本站总分。
怎样理解这份历史
每条记录对应来源快照中的一个原始指标。同一口径下相同值重复抓取不会制造新记录;数值、临时状态或来源状态变化时追加,核实后的方法版本另立指标身份。首个快照是本站的可追溯起点,不代表该指标第一次对外发布。当前模型来源为 Artificial Analysis。
Index、Elo、百分比等单位不能互相相减,也不会在这里合成为“综合总分”。
GDPval-AA 与 AA-Briefcase 的实时 Elo 现按裁判协议 v4.3.1 单独记录;旧记录的版本、值与日期保留,未区分裁判的旧值仅作历史参考。Intelligence Index 仍为 v4.3,官方用于指数的 Elo 贡献在模型加入时固定,不能用当前实时 Elo 重算指数。核对方法说明 ↗
APPEND-ONLY FACTS
观测记录
按观测时间倒序;同日多项指标分别保留。
| 观测日期 | 基准 / 版本 | 指标 | 原始值 | 来源 | 状态 |
|---|---|---|---|---|---|
| Artificial Analysis GDPval-AA任务 v2 · 裁判 v4.3.1实时 Elo · 裁判协议 v4.3.1 | Eloelo | 177 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3.1实时 Elo · 裁判协议 v4.3.1 | GDPval-AA v2elo | 177 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Model availabilitypoints | 6.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Model transparencypoints | 1.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Openness Indexindex | 38.9 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Posttraining data accesspoints | 0.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Posttraining data licensepoints | 0.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Pretraining data accesspoints | 0.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis Openness Indexvsource-unversioned | Pretraining data licensepoints | 0.0 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis GDPval-AAv2历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | Eloelo | 177 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3 | CritPtfraction | 0.6% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | GDPval-AA v2elo | 177 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3 | Humanity's Last Examfraction | 3.8% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis capability indicesv4.3 | Intelligence Indexindex | 8.9 | Artificial Analysis ↗第三方评测 | 临时 / 估算 | |
| Artificial Analysis intelligence component evaluationsv4.3 | AA-LCR v1.1fraction | 32.0% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3 | AA-Omniscienceindex | -19.7 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3 | AA-Omniscience accuracyfraction | 8.6% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.3 | AA-Omniscience hallucination ratefraction | 30.9% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 49.1 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 51.7 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 39.5 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 52.4 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | GDPval-AAelo | 230 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | GPQA Diamondfraction | 57.6% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | Humanity's Last Examfraction | 3.8% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Intelligence Indexindex | 12.2 | Artificial Analysis第三方评测 | 临时 / 估算 | |
| Artificial Analysis intelligence component evaluationsv4.1 | AA-LCRfraction | 33.0% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | AA-Omniscienceindex | -19.7 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 48.5 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 42.1 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 28.7 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | GDPval-AAelo | 230 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Intelligence Indexindex | 11.9 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | AA-Omniscienceindex | -19.6 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 89.7 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 23.1 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 28.1 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | GDPval-AAelo | 231 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | GPQA Diamondfraction | 52.2% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Intelligence Indexindex | 11.9 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 91.7 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 22.6 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Agentic Indexindex | 1.8 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Coding Indexindex | 9.4 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 27.6 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1历史参考 · 原记录未区分当前裁判协议,不作跨协议能力变化比较 | GDPval-AAelo | 231 | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | GDPval-AA normalizedfraction | 0.0% | Artificial Analysis ↗第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Intelligence Indexindex | 12.2 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 94.3 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | Tau Bankingfraction | 5.4% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | Terminal-Bench v2.1fraction | 1.9% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 22.3 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 上下文窗口tokens | 128,000 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | CritPtfraction | 0.6% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 端到端响应时间seconds | 27.5 秒 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | GPQA Diamondfraction | 57.6% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | Humanity's Last Examfraction | 3.7% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 输入价格 / 1M Tokenusd_per_1m_tokens | US$0.02 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis capability indicesvpublic-model-v1 | Intelligence Indexindex | 12.5 | Artificial Analysis第三方评测 | 临时 / 估算 | |
| Artificial Analysis intelligence component evaluationsv4.1 | AA-LCRfraction | 30.7% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | AA-Omniscienceindex | -20.1 | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 输出价格 / 1M Tokenusd_per_1m_tokens | US$0.1 | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 生成速度tokens_per_second | 94.6 Token/s | Model Coordinate repository snapshots第三方评测 | 有效 | |
| Artificial Analysis intelligence component evaluationsv4.1 | SciCodefraction | 24.4% | Artificial Analysis第三方评测 | 有效 | |
| Artificial Analysis model aggregatevpublic-model-v1 | 首个回答 Token 等待seconds | 22.3 秒 | Model Coordinate repository snapshots第三方评测 | 有效 |