Code review
结果只在同一分组内进行比较，因此用一种方式测试的模型绝不会与用另一种方式测试的模型同榜排名。
Review a code change and catch the real bugs it introduces without burying them in invalid or low-value comments
各排名分组使用独立的评分尺度。请只在同一分组内比较配置；EvalRank 绝不会把不同的系统压平到同一个榜单上。
跨基准共识
把该主题下已收录的基准合并成一份模型排名，权重与实际决策时一致。每个分数代表该模型成绩占最佳成绩的比例。
| 排名 | 模型 | 最佳占比 | 基准测试 |
|---|---|---|---|
| 1 | GPT-6.1 Terra | 100.0% | 1 |
| 2 | GPT-6.1 Sol | 90.5% | 1 |
| 3 | Grok 4.6 | 89.7% | 1 |
| 4 | Gemini 3.8 Flash | 87.1% | 1 |
| 5 | Gemini 4 Argon | 86.6% | 1 |
| 6 | Claude Opus 5.5 | 86.3% | 1 |
| 7 | GPT-6 Sol | 85.2% | 1 |
| 8 | Claude Opus 5 | 80.7% | 1 |
| 9 | GPT-6 Luna | 76.8% | 1 |
智能体系统 · 智能体式
探索性——尚未产生排名冠军rg-code-review-agent-system-agentic-agent-system-v1Explorer 依据
MacroscopeBench, every reasoning effort (Macroscope)
各来源不一致，分别展示，绝不取平均。观测于 ；有效至 。
| 排名 | 配置 | 能力 | 不确定性 | 家族 |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 macroscopebench:max-effortconfig_e6d73883b79a458bb42954a12f9a4bd2f61b19d1a69dbedddcee57a7ccfc3a20 | 100% | 未知 | 1 |
| 2 | GPT-6.1 Terra macroscopebench:high-effortconfig_0fca9081059f9ea84ea730e704c71ba4a674756223e8349e7fb8c727c9279e18 | 96% | 未知 | 1 |
| 3 | GPT-6.1 Terra macroscopebench:medium-effortconfig_c3d1dc0e44423d43a98bc671e568d383d8cef0d30029eb2e6b34adb2bb3ffb5e | 93% | 未知 | 1 |
| 4 | GPT-6 Sol macroscopebench:max-effortconfig_8355540786508f62b7649c7e41ab7c7c3840eb783dc6fa18f30fa4ebf7f69482 | 91% | 未知 | 1 |
| 5 | Claude Opus 5.5 macroscopebench:xhigh-effortconfig_8b30978b3842998bf9bbb38614b78c91c8e326d7807192a9af2c2988cda4e00f | 87% | 未知 | 1 |
| 6 | GPT-6.1 Sol macroscopebench:max-effortconfig_78df30b801e7eb04d6ef351803a5b54130f7c6a7b4d5ff9aa9756422418d076d | 86% | 未知 | 1 |
| 7 | GPT-6 Astra macroscopebench:max-effortconfig_a75d727f10d4a0eac19d6e01bf487c6184e8a510672183ea02a01a5fc23f3c8f | 85% | 未知 | 1 |
| 8 | GPT-5.6 Sol macroscopebench:high-effortconfig_99f8f7f0f456b8c6d1df1fa7043c58df55379cef41fa8a486e932ad43c3dfaca | 84% | 未知 | 1 |
| 9 | GLM 5.3 macroscopebench:max-effortconfig_483032ccc1d3f89ac4535ef6e623763c1cf3e47b0f87f7636ef89ca557e32604 | 83% | 未知 | 1 |
| 10 | GPT-6.1 Sol macroscopebench:xhigh-effortconfig_82926d8185ff4aefa885e0248125762b7364ee16ea39e3de6323d76c6420dbfb | 83% | 未知 | 1 |
| 11 | GPT-6 Sol macroscopebench:xhigh-effortconfig_b5112dbc3d69b149f10e0a3d43ec90644a4a03cb6431d7441b6cd137f4d727e3 | 83% | 未知 | 1 |
| 12 | GPT-6.1 Sol macroscopebench:high-effortconfig_5a24f76c989220b98e8d8a772ebe10c60ec362db74a637fec8cbe6db7ddc6300 | 82% | 未知 | 1 |
| 13 | GPT-6.1 Terra macroscopebench:low-effortconfig_91304552afd01f5c37cfc14bc2b27d07892f9db62ffe0fbff8d4206288f68ae8 | 80% | 未知 | 1 |
| 14 | Grok 4.6 macroscopebench:xhigh-effortconfig_40f36156eaa01fa3d0fa03fe65b5d085cd5e91e64eeff83efcb02980d3ba0628 | 80% | 未知 | 1 |
| 15 | Grok 4.6 macroscopebench:high-effortconfig_9c445642dfb058b00be7b564987b4f27af2bff78275a8d43bda504fd245abf43 | 78% | 未知 | 1 |
| 16 | GPT-6 Sol macroscopebench:high-effortconfig_e25f44e408a733401428874642e5edfe2bf935f05a41052f7ab74ee0fbb99318 | 76% | 未知 | 1 |
| 17 | GPT-6.1 Sol macroscopebench:medium-effortconfig_9dafacf26cdd886ba1c1f49d87065b7d3f5cd54feb06f8ad8d6638188d37d663 | 73% | 未知 | 1 |
| 18 | Claude Opus 5.5 macroscopebench:high-effortconfig_733dd5cc9437841e5449fa06b422c17e83334f55168c2f5d5b9325a535289ffc | 72% | 未知 | 1 |
| 19 | Kimi K3 macroscopebench:max-effortconfig_fc2b52a594c8a99a3b7c2b59e1391c3c83620d834ea81943588d979eda112be6 | 72% | 未知 | 1 |
| 20 | Grok 4.6 macroscopebench:medium-effortconfig_10f0ba0fa4acbaba4d1227648e8a00f2cb7bc77f44f64a7b4ccf1880e161daf2 | 72% | 未知 | 1 |
| 21 | Gemini 4 Argon macroscopebench:high-effortconfig_530c748fedff86cf7de491f4118d6c8a2c4dd2493b47a2cced04258747e2abf5 | 72% | 未知 | 1 |
| 22 | GPT-6 Luna macroscopebench:max-effortconfig_3c0dd5fe33609fcaddb22f06a3e078e4f268b2003568cc116ae7eb7384151607 | 71% | 未知 | 1 |
| 23 | DeepSeek V4.1 Flash macroscopebench:max-effortconfig_1b29e838bc883e7f200532f0c15b9414fcb38a2e33e878b81ffa9203b1f651e2 | 70% | 未知 | 1 |
| 24 | Gemini 3.8 Flash macroscopebench:medium-effortconfig_9da81bf684efbe1ad1948d51b5b78cb6cd93e9dad9c50d8fc7887883e3a32b50 | 66% | 未知 | 1 |
| 25 | DeepSeek V4.1 Flash macroscopebench:low-effortconfig_ce17896acd252f51af585df90ee2ec0e3d45888ceb07c6a1dd33a711252d6ed4 | 66% | 未知 | 1 |
| 26 | GPT-6.1 Sol macroscopebench:low-effortconfig_7c8deb8f693e6c268a2313d787bf120b0ea2e5f62cd4f1417060353ee57317f8 | 66% | 未知 | 1 |
| 27 | DeepSeek V4.1 Flash macroscopebench:high-effortconfig_2b2e93dfccd56f89985469205995a700afa68ca234cb33876dd8bebc4d1b4b49 | 66% | 未知 | 1 |
| 28 | Gemini 4 Argon macroscopebench:medium-effortconfig_d54438a864b253c5bd74a5728f3089cefd93808fcbd5b5037bffb483fbb03af1 | 65% | 未知 | 1 |
| 29 | Claude Opus 5.5 macroscopebench:medium-effortconfig_81b476353cd67c47751bf44dac77ad0f9eee84f3498f3c8b89f694420f430ff3 | 65% | 未知 | 1 |
| 30 | Claude Opus 5 macroscopebench:high-effortconfig_88f804bb21b203128d39a5f2c65aac58f46eaf269947654a8c16b50f2d1a8219 | 64% | 未知 | 1 |
| 31 | GLM 5.3 macroscopebench:high-effortconfig_f8c73c8a27ed8ec1ddba264606a44cea6f03daac9b87b51d8be381f565cbeb04 | 63% | 未知 | 1 |
| 32 | GPT-6 Sol macroscopebench:medium-effortconfig_60c27177bb62cbb0bc1a1315c3195c7739acd13b0107510df793d173238ae34f | 63% | 未知 | 1 |
| 33 | Kimi K3 macroscopebench:high-effortconfig_9197026ae1add69be5447726f323c3b6df043eaf50ff17785efd62297e0def15 | 61% | 未知 | 1 |
| 34 | GPT-6 Luna macroscopebench:xhigh-effortconfig_88e3c4fc4945be6778e224c4012ab03451b10e72c22e0ee09ea2182d94eb2703 | 60% | 未知 | 1 |
| 35 | GPT-5.6 Sol macroscopebench:low-effortconfig_accd884ef51fabc161a52d15d9b4db38cca243783d42ed7108a2f8b5d73a3402 | 55% | 未知 | 1 |
| 36 | GPT-6 Luna macroscopebench:high-effortconfig_82058abb9bbc34aa1c34ae904cca0e264290b8f3b3f4e00bd6a94f4416e17652 | 53% | 未知 | 1 |
| 37 | Claude Opus 5 macroscopebench:medium-effortconfig_88aafab5a5dd6c8836dce41a23b99df08d5a0740972946670a44b907435db23b | 53% | 未知 | 1 |
| 38 | Muse Wonder macroscopebench:low-effortconfig_56cc703ca619ffb3e91db893245ecd9cbcce726abcaf09352122dfdf63726f56 | 50% | 未知 | 1 |
| 39 | GPT-6 Luna macroscopebench:medium-effortconfig_29792f84f3e1cb2bc35c50d8e07675f7a7d08a8e1772251c5544842835e305ba | 45% | 未知 | 1 |
| 40 | GPT-6 Sol macroscopebench:low-effortconfig_ef2e7cca45952dfa1c5f5bb88e5a50eb9988bf57d5be8e41595685154337e61f | 44% | 未知 | 1 |
| 41 | Claude Opus 5.5 macroscopebench:low-effortconfig_0c2ade372c70318cb921a18ade2cfc1572dc5691bf9b7db84d5b9bb1f4c19160 | 43% | 未知 | 1 |
| 42 | Gemini 4 Argon macroscopebench:low-effortconfig_9f37b46f7c7caa07013c896b1c888f7c6cc019824df18eb87c248740acd08540 | 42% | 未知 | 1 |
| 43 | Claude Opus 5 macroscopebench:low-effortconfig_db91574819230d1677188be05a134058093d98912c162f534aff9147cb2b76e9 | 38% | 未知 | 1 |
| 44 | Grok 4.6 macroscopebench:low-effortconfig_3368cc140fbfd7ce4dadd75a37bb9e79ffa5019153bd0554318a22c185f860cb | 32% | 未知 | 1 |
| 45 | GLM 5.3 macroscopebench:low-effortconfig_888ee1fd139c4c552ceab0300b8ab5cd4d374e6a23f8b8f36006393e328f593b | 25% | 未知 | 1 |
| 46 | Kimi K3 macroscopebench:low-effortconfig_7aab349be7a0f827ef808b8c22a1e74af5ea0f2d625d7b523b9fb39ed3d32db7 | 22% | 未知 | 1 |
| 47 | Gemini 3.8 Flash macroscopebench:low-effortconfig_02e963f4bb6cd6a43724fd3a7e10a67dc2f157f877b244978c1d858c2e24d18d | 17% | 未知 | 1 |
| 48 | GPT-6 Luna macroscopebench:low-effortconfig_246e3f618910cca9f31cb112d2ec39ccdad228a931a1c22d0b5fd7430147faaf | 0% | 未知 | 1 |
来源：Macroscope
Explorer 依据
MacroscopeBench, medium effort (Macroscope)
各来源不一致，分别展示，绝不取平均。观测于 ；有效至 。
| 排名 | 配置 | 能力 | 不确定性 | 家族 |
|---|---|---|---|---|
| 1 | GPT-6.1 Terra macroscopebench:medium-effortconfig_c3d1dc0e44423d43a98bc671e568d383d8cef0d30029eb2e6b34adb2bb3ffb5e | 100% | 未知 | 1 |
| 2 | GPT-6.1 Sol macroscopebench:medium-effortconfig_9dafacf26cdd886ba1c1f49d87065b7d3f5cd54feb06f8ad8d6638188d37d663 | 59% | 未知 | 1 |
| 3 | Grok 4.6 macroscopebench:medium-effortconfig_10f0ba0fa4acbaba4d1227648e8a00f2cb7bc77f44f64a7b4ccf1880e161daf2 | 55% | 未知 | 1 |
| 4 | Gemini 3.8 Flash macroscopebench:medium-effortconfig_9da81bf684efbe1ad1948d51b5b78cb6cd93e9dad9c50d8fc7887883e3a32b50 | 44% | 未知 | 1 |
| 5 | Gemini 4 Argon macroscopebench:medium-effortconfig_d54438a864b253c5bd74a5728f3089cefd93808fcbd5b5037bffb483fbb03af1 | 42% | 未知 | 1 |
| 6 | Claude Opus 5.5 macroscopebench:medium-effortconfig_81b476353cd67c47751bf44dac77ad0f9eee84f3498f3c8b89f694420f430ff3 | 41% | 未知 | 1 |
| 7 | GPT-6 Sol macroscopebench:medium-effortconfig_60c27177bb62cbb0bc1a1315c3195c7739acd13b0107510df793d173238ae34f | 36% | 未知 | 1 |
| 8 | Claude Opus 5 macroscopebench:medium-effortconfig_88aafab5a5dd6c8836dce41a23b99df08d5a0740972946670a44b907435db23b | 16% | 未知 | 1 |
| 9 | GPT-6 Luna macroscopebench:medium-effortconfig_29792f84f3e1cb2bc35c50d8e07675f7a7d08a8e1772251c5544842835e305ba | 0% | 未知 | 1 |
来源：Macroscope
独立依据不足，无法形成头部梯队
- 校准尚未通过验证
- 独立基准族的覆盖度低于要求的门槛