Content-safety classification

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。

Classify text as safe or unsafe, flagging unsafe content with high recall while rarely flagging harmless text that only sounds unsafe

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	nemotron3ultra550ba55b	100.0%	1
2	gemini31flashlite	99.7%	1
3	beam10501ba23b	99.6%	1
4	gemini31flashlitepreview	99.5%	1
5	gpt61solpro	99.3%	1
6	gemini3flashpreview	99.1%	1
7	claudesonnet55	99.1%	1
8	minimaxm3	99.0%	1
9	gpt61terra	98.9%	1
10	minimaxm27	98.9%	1
11	minimaxm25	98.8%	1
12	step5preview	98.8%	1
13	gpt6sol	98.8%	1
14	gemma431bit	98.6%	1
15	gemma426ba4bit	98.5%	1
16	gpt61sol	98.5%	1
17	kimik26	98.4%	1
18	gpt6solpro	98.4%	1
19	musesparkopen	98.3%	1
20	glimpse	98.2%	1
21	qwen35397ba17b	98.2%	1
22	nemotron35supervl	98.1%	1
23	deepseekv41flash1017	97.9%	1
24	minimaxm31flash	97.9%	1
25	deepseekv4flashvisionexp	97.8%	1
26	gemini35flash	97.6%	1
27	gpt61terrapro	97.4%	1
28	mimov26flash	97.4%	1
29	museglimmer30b	97.3%	1
30	deepseekv4pro0813	97.3%	1
31	glm52	97.2%	1
32	glm47	96.9%	1
33	gemini36flash	96.9%	1
34	gemini4argon	96.8%	1
35	deepseekv32	96.8%	1
36	qwen3527b	96.8%	1
37	gemini35flashlite	96.6%	1
38	claudehaiku55	96.6%	1
39	deepseekv41flash	96.6%	1
40	glm53flash	96.5%	1
41	gpt51	96.4%	1
42	qwen3627b	96.1%	1
43	deepseekv4pro	95.9%	1
44	glm53	95.4%	1
45	qwen3827b	95.4%	1
46	gpt6lunapro	95.3%	1
47	granite428b	95.1%	1
48	gpt56lunapro	95.0%	1
49	nemotron3super120ba12b	94.9%	1
50	qwen3635ba3b	94.9%	1
51	deepseekv4flash	94.8%	1
52	gpt6luna	94.6%	1
53	gpt54mini	94.6%	1
54	gpt56luna	94.4%	1
55	deepseekv4flash0731	92.4%	1
56	glm47flash	92.0%	1
57	gpt51codex	92.0%	1
58	gpt51codexmini	91.5%	1
59	glm5	90.3%	1
60	nemotron35lightning	90.1%	1
61	gpt51codexmax	89.8%	1
62	gpt54nano	88.2%	1
63	claudesonnet5	88.1%	1
64	qwen3535ba3b	80.5%	1

模型配置 · 直接提示

探索性——尚未产生排名冠军

rg-content-safety-classification-model-configuration-direct-prompt-model-configuration-v1

Explorer 依据

Content-safety classification board (EvalRank)

解析时间：昨天

单一来源。观测于 2026-11-03T00:01:18Z；有效至 2027-02-01T00:01:18Z。

排名	配置	能力	不确定性	家族
1	nvidia/nemotron-3-ultra-550b-a55b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_7d99eb067bf755ad2ac4eed492ded7c298113a972ed3482133ed762b1de16b78`	100%	未知	1
2	google/gemini-3.1-flash-lite evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_c0740610436eb8ba7a4c51adfc38037e58a5793315d734e1937878f4649d3caf`	98%	未知	1
3	reflection/beam-1.0-501b-a23b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_2155c6e9341ea1063997ef1c246a68b92344920065b206e8a532d5dd4039fb28`	98%	未知	1
4	google/gemini-3.1-flash-lite-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_4dd62c9832f9c5af54479fb9a4ab1798867614257d96152e7c372520ac11f083`	98%	未知	1
5	openai/gpt-6.1-sol-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_c3efa144559b10cd3f0dbd1f2744d19cd8f6f8744fcad431a5c8737bd883ba80`	96%	未知	1
6	google/gemini-3-flash-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_442b3254e105769c9f1fbafa2cf80ca2cf15b6eaa49b5a6befcc7744db4a57a7`	96%	未知	1
7	anthropic/claude-sonnet-5.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_9b18ddbe788f23e2328627ca091bbfa2b87facb463b3abe9156f460009535581`	96%	未知	1
8	minimax/minimax-m3 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_fab1fceeb9494c0411fa9c92a0eaba632dc8409c976d16c209063dfe93ccdc29`	95%	未知	1
9	openai/gpt-6.1-terra evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_1d22c134a1360bbfc2ae951d071b1056af11d422a2ba518bd8916e80994a3c01`	94%	未知	1
10	minimax/minimax-m2.7 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_b23ed36c7637c457294b272eadb80afcead7d23f861e543d858fa7bdefd5b7d0`	94%	未知	1
11	minimax/minimax-m2.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_ddc9b3a2df0fb875265ef3ed132304ba7acc702751b37091056e002a5f1329f9`	94%	未知	1
12	stepfun/step-5-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_ea728c9b0fca06110f24790229ec3a0e2a90693fcaca272afd50c4efa4aa8bf4`	94%	未知	1
13	openai/gpt-6-sol evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_fad8a2e4ecc1874a0f01ab8ef3a739d8b2384ea466ef8be154432a9a64592ad5`	94%	未知	1
14	google/gemma-4-31b-it evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_6f2bca52853c147b75881629111e70f00705ce89cdadd157111b158d296f5b7a`	93%	未知	1
15	google/gemma-4-26b-a4b-it evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_43b6f57ea4f493c88995b41322c8d8e3e3b2cf02f67b9f56fc53750fbeecf594`	93%	未知	1
16	openai/gpt-6.1-sol evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_a0dc3da2feef4024d38c3593705754bd8fd4590873de9806a763f792bb236ddb`	92%	未知	1
17	moonshotai/kimi-k2.6 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_635e37df07b4bda8605aafb0b8e7b6a71187e5b293bece532dbc023f07fdc1b9`	92%	未知	1
18	openai/gpt-6-sol-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_8fb82d4ae23949fa0e3b09224a701932e00db56e372b300563f6fc9696041d81`	92%	未知	1
19	meta/muse-spark-open evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_7f227ac4a9e1b2c8f6e3f78a6434486ff5eaa964afe06e1ab7ef7d7141209bbd`	91%	未知	1
20	elorian/glimpse evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_238f80b31c4c82b5210bf16843248970293dbaa425171d3c390e1d1cd5393d8a`	91%	未知	1
21	qwen/qwen3.5-397b-a17b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_ad0d0fa0e3daf5c7f909972ba32614dbea12f65f4db742e32841b08fa389f2f3`	91%	未知	1
22	nvidia/nemotron-3.5-super-vl evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_115f5e3115fa9e1af3e6d8dc67b0b1a66b9cacaf359a50758cda8b35ef7a61d2`	90%	未知	1
23	deepseek/deepseek-v4.1-flash-1017 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_3dd050f56d024ad6316e2ea453b46bc3d39832b2aaf9abd34cf8e5e7ba34cf1f`	89%	未知	1
24	minimax/minimax-m3.1-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_a5affc900ec149755655e39d490eac5d2e85880361ddf9f5a97becd6044e7964`	89%	未知	1
25	deepseek/deepseek-v4-flash-vision-exp evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_3d2d55c4242f8eb51b7b133145e82ef33971385ece7ee7fd0abc0f333643c6b1`	89%	未知	1
26	google/gemini-3.5-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_8610441536c049d3ee484d51228f0171f9aa0d86ede49e6031c3aac345c19412`	88%	未知	1
27	openai/gpt-6.1-terra-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_73f0264984f0b9978d96b597d40a7268f75362e2212bb641b5792a45d3644cf0`	87%	未知	1
28	xiaomi/mimo-v2.6-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_f4356f31a8b09b0ac18d413260140e3d740e4e498bf5b2d6e81d26348691a909`	86%	未知	1
29	meta/muse-glimmer-30b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_6747ee32ad5046a79edfd2e3b261add7a9c324422df989ce7ed8f096f0ee3c28`	86%	未知	1
30	deepseek/deepseek-v4-pro-0813 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_f9b97ce14e997236c72e6daade92cde3c60dac1f0ba7c984c6669fc68f7e9c1d`	86%	未知	1
31	z-ai/glm-5.2 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_120dd6c22755d4cbdc9272d709391f21fd17a896d825e04a03e75f2f1843a560`	86%	未知	1
32	z-ai/glm-4.7 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_64619a046266e3fc27a85eb007cea8df376927c7f5df6177bbf9dff0d0e9a3aa`	84%	未知	1
33	google/gemini-3.6-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_5f1804314139414d4217c30c800f66773418cc18c9daf8252c03693931fa9470`	84%	未知	1
34	google/gemini-4-argon evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_f08e4ef4677affaf9bcf7fa8ad339cb268daa2169c1395f91ccdc3848dc38fa1`	84%	未知	1
35	deepseek/deepseek-v3.2 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:unavailable`config_e46f182963294fd8ac9b13dd94e005481af49bb73d57f6b4c1ff58765f7ff36d`	83%	未知	1
36	qwen/qwen3.5-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_59baf54cd1ce90bf480c5f8f6be017b9bcd32182817e84291a68b45e926a7275`	83%	未知	1
37	google/gemini-3.5-flash-lite evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_8ab35556769122f6730cb555f916447e1c8bf6185e10523f28b3ce5e50db27fb`	83%	未知	1
38	anthropic/claude-haiku-5.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_2751fa6baf3817144e1312d65bc49b600a27ff9796b40b892cf5ed09e6b00568`	83%	未知	1
39	deepseek/deepseek-v4.1-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_1d325a8aeb593573cca9ba825520046da0ffc4fcefcb06db061f8ef5fea91a54`	82%	未知	1
40	z-ai/glm-5.3-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_fb52ad4ac1c4026292a3f1240a4dc2e1d12e16083e95646db8212b77ed57ea1b`	82%	未知	1
41	openai/gpt-5.1 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_b8350fce891d7d514e0480df7b6234e491f02dbc8ae150fd6e704aa23b0191e6`	82%	未知	1
42	qwen/qwen3.6-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_a78ddb6cebf92f398affb21e6b7e3f795bc29270d9fd53deead33eaa1df411cb`	80%	未知	1
43	deepseek/deepseek-v4-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_7000fdb416cbab9f2be6b2f33119adc41118915abc07e7a7e8213ee7336d205a`	79%	未知	1
44	z-ai/glm-5.3 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_12fe0ff251166fdeb583bf45c9ece4a19c1d3bb40b7c33662ba2c969d9850eaf`	76%	未知	1
45	qwen/qwen3.8-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_be71109b07ef89f80a58dc5652e8da5649a32fa622ccbe7af6a01c49b662325f`	76%	未知	1
46	openai/gpt-6-luna-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_3b22b0a8f90d6a9c89cb944c4ada16295394fed726da9f02edcefa4b6454fb38`	76%	未知	1
47	ibm-granite/granite-4.2-8b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_4c77c6e59730f431caf2e6018aa101251b026a28699cca491f6fcb57d134a639`	75%	未知	1
48	openai/gpt-5.6-luna-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_a84bda106ab3974d69df8143a7ecfd41ae25fa3af4034c9cda5232a494ed381d`	74%	未知	1
49	nvidia/nemotron-3-super-120b-a12b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_65206ed835937e4a58e9249a2680b4d2872397bf3abf72b22d2c07dc044ed271`	74%	未知	1
50	qwen/qwen3.6-35b-a3b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_099dfd7875cbd5ab4e85221be7656ef51863c17bb94c9af651a328e19a70a2e8`	74%	未知	1
51	deepseek/deepseek-v4-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_6bd0c5eb925bb5e63bef71def75efe0301ca52099eca2e0bbcf7981e2aceb51c`	73%	未知	1
52	openai/gpt-6-luna evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_4dbbd2f0ab0d84b7c248fbce5e80ca7551aad370b42a727b4f69166c20266440`	73%	未知	1
53	openai/gpt-5.4-mini evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_4b66200184124eb485b499c0dd56642e2430b644289885afea04aa641c22cda7`	72%	未知	1
54	openai/gpt-5.6-luna evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_5cde82caefa8a815fe6f4beb950c65f9fae31efceeda19d1f631601ccdd4b6d3`	72%	未知	1
55	deepseek/deepseek-v4-flash-0731 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_e28e0dd0460ebad70a0f394de212558dcc7fb3d62e127c067a94eea4d50e7184`	61%	未知	1
56	z-ai/glm-4.7-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_79cff7558daca1c8edbb265f6a708011a390a18f0d07f3295f2944d16c652731`	59%	未知	1
57	openai/gpt-5.1-codex evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_843a8c73b4ee4d1e2dca77b266a2d0a8c20288a24b00117aa916efad91b95622`	59%	未知	1
58	openai/gpt-5.1-codex-mini evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_7791eec96f3a1571763eb0485f92474fc5c2326ef8138f86f49d13bdc8b9a5de`	57%	未知	1
59	z-ai/glm-5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_d92561080dcb2ec18134f028e6d589c34c5ff86933ce31cd0baef9e792810470`	50%	未知	1
60	nvidia/nemotron-3.5-lightning evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_0d55bafe2f486b6d40691b51a77137b641a3498419487b57ed2952d7d09a3483`	49%	未知	1
61	openai/gpt-5.1-codex-max evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_low`config_7af97febe265676b35c339b4b079bd74568d699f56190ed97a84580bfe2c2ffb`	47%	未知	1
62	openai/gpt-5.4-nano evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_0ab43c9a147fa575ea2d61319dfac2869684ccb48595d6ba5ef34cca75dc9286`	40%	未知	1
63	anthropic/claude-sonnet-5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_962021295d532bdfe74c91e367fbb71a3bdb956d6247cd8302ccca6d864db388`	39%	未知	1
64	qwen/qwen3.5-35b-a3b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_off`config_8d6e5890ab2b84a55f734f9f18f5b5329413ddfa75155e2ecb7b4bd460f6dcf7`	0%	未知	1

对比该分组中的配置

evalrank-content-safety benchmark evidence

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

返回全部工作类别