Content-safety classification

I risultati vengono confrontati solo all'interno dello stesso gruppo, così un modello testato in un modo non viene mai messo in classifica contro un modello testato in un altro.

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

I gruppi di classifica usano scale diverse. Confronta le configurazioni solo all'interno dello stesso gruppo; EvalRank non appiattisce mai sistemi diversi in un'unica classifica.

Consenso tra i benchmark

Una classifica dei modelli sui benchmark ammessi per questo argomento, con lo stesso peso che si dà alle decisioni. Ogni punteggio è la quota del miglior risultato ottenuta dal modello.

PosizioneModelloQuota del miglioreBenchmark
1nemotron3ultra550ba55b100.0%1
2gemini31flashlite99.7%1
3beam10501ba23b99.6%1
4gemini31flashlitepreview99.5%1
5gpt61solpro99.3%1
6gemini3flashpreview99.1%1
7claudesonnet5599.1%1
8minimaxm399.0%1
9gpt61terra98.9%1
10minimaxm2798.9%1
11minimaxm2598.8%1
12step5preview98.8%1
13gpt6sol98.8%1
14gemma431bit98.6%1
15gemma426ba4bit98.5%1
16gpt61sol98.5%1
17kimik2698.4%1
18gpt6solpro98.4%1
19musesparkopen98.3%1
20glimpse98.2%1
21qwen35397ba17b98.2%1
22nemotron35supervl98.1%1
23deepseekv41flash101797.9%1
24minimaxm31flash97.9%1
25deepseekv4flashvisionexp97.8%1
26gemini35flash97.6%1
27gpt61terrapro97.4%1
28mimov26flash97.4%1
29museglimmer30b97.3%1
30deepseekv4pro081397.3%1
31glm5297.2%1
32glm4796.9%1
33gemini36flash96.9%1
34gemini4argon96.8%1
35deepseekv3296.8%1
36qwen3527b96.8%1
37gemini35flashlite96.6%1
38claudehaiku5596.6%1
39deepseekv41flash96.6%1
40glm53flash96.5%1
41gpt5196.4%1
42qwen3627b96.1%1
43deepseekv4pro95.9%1
44glm5395.4%1
45qwen3827b95.4%1
46gpt6lunapro95.3%1
47granite428b95.1%1
48gpt56lunapro95.0%1
49nemotron3super120ba12b94.9%1
50qwen3635ba3b94.9%1
51deepseekv4flash94.8%1
52gpt6luna94.6%1
53gpt54mini94.6%1
54gpt56luna94.4%1
55deepseekv4flash073192.4%1
56glm47flash92.0%1
57gpt51codex92.0%1
58gpt51codexmini91.5%1
59glm590.3%1
60nemotron35lightning90.1%1
61gpt51codexmax89.8%1
62gpt54nano88.2%1
63claudesonnet588.1%1
64qwen3535ba3b80.5%1

Configurazione del modello · Prompt diretto

Esplorativo - non è ancora un vincitore in classifica
rg-content-safety-classification-model-configuration-direct-prompt-model-configuration-v1

Evidenze dell'explorer

Content-safety classification board (EvalRank)

Fonte unica. Osservato il ; valido fino al .

PosizioneConfigurazioneCapacitàIncertezzaFamiglie
1nvidia/nemotron-3-ultra-550b-a55b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_7d99eb067bf755ad2ac4eed492ded7c298113a972ed3482133ed762b1de16b78100%Sconosciuto1
2google/gemini-3.1-flash-lite evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_c0740610436eb8ba7a4c51adfc38037e58a5793315d734e1937878f4649d3caf98%Sconosciuto1
3reflection/beam-1.0-501b-a23b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_2155c6e9341ea1063997ef1c246a68b92344920065b206e8a532d5dd4039fb2898%Sconosciuto1
4google/gemini-3.1-flash-lite-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_4dd62c9832f9c5af54479fb9a4ab1798867614257d96152e7c372520ac11f08398%Sconosciuto1
5openai/gpt-6.1-sol-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_c3efa144559b10cd3f0dbd1f2744d19cd8f6f8744fcad431a5c8737bd883ba8096%Sconosciuto1
6google/gemini-3-flash-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_442b3254e105769c9f1fbafa2cf80ca2cf15b6eaa49b5a6befcc7744db4a57a796%Sconosciuto1
7anthropic/claude-sonnet-5.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_9b18ddbe788f23e2328627ca091bbfa2b87facb463b3abe9156f46000953558196%Sconosciuto1
8minimax/minimax-m3 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_fab1fceeb9494c0411fa9c92a0eaba632dc8409c976d16c209063dfe93ccdc2995%Sconosciuto1
9openai/gpt-6.1-terra evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_1d22c134a1360bbfc2ae951d071b1056af11d422a2ba518bd8916e80994a3c0194%Sconosciuto1
10minimax/minimax-m2.7 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_b23ed36c7637c457294b272eadb80afcead7d23f861e543d858fa7bdefd5b7d094%Sconosciuto1
11minimax/minimax-m2.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_ddc9b3a2df0fb875265ef3ed132304ba7acc702751b37091056e002a5f1329f994%Sconosciuto1
12stepfun/step-5-preview evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_ea728c9b0fca06110f24790229ec3a0e2a90693fcaca272afd50c4efa4aa8bf494%Sconosciuto1
13openai/gpt-6-sol evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_fad8a2e4ecc1874a0f01ab8ef3a739d8b2384ea466ef8be154432a9a64592ad594%Sconosciuto1
14google/gemma-4-31b-it evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_6f2bca52853c147b75881629111e70f00705ce89cdadd157111b158d296f5b7a93%Sconosciuto1
15google/gemma-4-26b-a4b-it evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_43b6f57ea4f493c88995b41322c8d8e3e3b2cf02f67b9f56fc53750fbeecf59493%Sconosciuto1
16openai/gpt-6.1-sol evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_a0dc3da2feef4024d38c3593705754bd8fd4590873de9806a763f792bb236ddb92%Sconosciuto1
17moonshotai/kimi-k2.6 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_635e37df07b4bda8605aafb0b8e7b6a71187e5b293bece532dbc023f07fdc1b992%Sconosciuto1
18openai/gpt-6-sol-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_8fb82d4ae23949fa0e3b09224a701932e00db56e372b300563f6fc9696041d8192%Sconosciuto1
19meta/muse-spark-open evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_7f227ac4a9e1b2c8f6e3f78a6434486ff5eaa964afe06e1ab7ef7d7141209bbd91%Sconosciuto1
20elorian/glimpse evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_238f80b31c4c82b5210bf16843248970293dbaa425171d3c390e1d1cd5393d8a91%Sconosciuto1
21qwen/qwen3.5-397b-a17b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_ad0d0fa0e3daf5c7f909972ba32614dbea12f65f4db742e32841b08fa389f2f391%Sconosciuto1
22nvidia/nemotron-3.5-super-vl evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_115f5e3115fa9e1af3e6d8dc67b0b1a66b9cacaf359a50758cda8b35ef7a61d290%Sconosciuto1
23deepseek/deepseek-v4.1-flash-1017 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_3dd050f56d024ad6316e2ea453b46bc3d39832b2aaf9abd34cf8e5e7ba34cf1f89%Sconosciuto1
24minimax/minimax-m3.1-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_a5affc900ec149755655e39d490eac5d2e85880361ddf9f5a97becd6044e796489%Sconosciuto1
25deepseek/deepseek-v4-flash-vision-exp evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_3d2d55c4242f8eb51b7b133145e82ef33971385ece7ee7fd0abc0f333643c6b189%Sconosciuto1
26google/gemini-3.5-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_8610441536c049d3ee484d51228f0171f9aa0d86ede49e6031c3aac345c1941288%Sconosciuto1
27openai/gpt-6.1-terra-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_73f0264984f0b9978d96b597d40a7268f75362e2212bb641b5792a45d3644cf087%Sconosciuto1
28xiaomi/mimo-v2.6-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_f4356f31a8b09b0ac18d413260140e3d740e4e498bf5b2d6e81d26348691a90986%Sconosciuto1
29meta/muse-glimmer-30b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_6747ee32ad5046a79edfd2e3b261add7a9c324422df989ce7ed8f096f0ee3c2886%Sconosciuto1
30deepseek/deepseek-v4-pro-0813 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_f9b97ce14e997236c72e6daade92cde3c60dac1f0ba7c984c6669fc68f7e9c1d86%Sconosciuto1
31z-ai/glm-5.2 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_120dd6c22755d4cbdc9272d709391f21fd17a896d825e04a03e75f2f1843a56086%Sconosciuto1
32z-ai/glm-4.7 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_64619a046266e3fc27a85eb007cea8df376927c7f5df6177bbf9dff0d0e9a3aa84%Sconosciuto1
33google/gemini-3.6-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_5f1804314139414d4217c30c800f66773418cc18c9daf8252c03693931fa947084%Sconosciuto1
34google/gemini-4-argon evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_f08e4ef4677affaf9bcf7fa8ad339cb268daa2169c1395f91ccdc3848dc38fa184%Sconosciuto1
35deepseek/deepseek-v3.2 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:unavailableconfig_e46f182963294fd8ac9b13dd94e005481af49bb73d57f6b4c1ff58765f7ff36d83%Sconosciuto1
36qwen/qwen3.5-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_59baf54cd1ce90bf480c5f8f6be017b9bcd32182817e84291a68b45e926a727583%Sconosciuto1
37google/gemini-3.5-flash-lite evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_8ab35556769122f6730cb555f916447e1c8bf6185e10523f28b3ce5e50db27fb83%Sconosciuto1
38anthropic/claude-haiku-5.5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_2751fa6baf3817144e1312d65bc49b600a27ff9796b40b892cf5ed09e6b0056883%Sconosciuto1
39deepseek/deepseek-v4.1-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_1d325a8aeb593573cca9ba825520046da0ffc4fcefcb06db061f8ef5fea91a5482%Sconosciuto1
40z-ai/glm-5.3-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_fb52ad4ac1c4026292a3f1240a4dc2e1d12e16083e95646db8212b77ed57ea1b82%Sconosciuto1
41openai/gpt-5.1 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_b8350fce891d7d514e0480df7b6234e491f02dbc8ae150fd6e704aa23b0191e682%Sconosciuto1
42qwen/qwen3.6-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_a78ddb6cebf92f398affb21e6b7e3f795bc29270d9fd53deead33eaa1df411cb80%Sconosciuto1
43deepseek/deepseek-v4-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_7000fdb416cbab9f2be6b2f33119adc41118915abc07e7a7e8213ee7336d205a79%Sconosciuto1
44z-ai/glm-5.3 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_12fe0ff251166fdeb583bf45c9ece4a19c1d3bb40b7c33662ba2c969d9850eaf76%Sconosciuto1
45qwen/qwen3.8-27b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_be71109b07ef89f80a58dc5652e8da5649a32fa622ccbe7af6a01c49b662325f76%Sconosciuto1
46openai/gpt-6-luna-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_3b22b0a8f90d6a9c89cb944c4ada16295394fed726da9f02edcefa4b6454fb3876%Sconosciuto1
47ibm-granite/granite-4.2-8b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_4c77c6e59730f431caf2e6018aa101251b026a28699cca491f6fcb57d134a63975%Sconosciuto1
48openai/gpt-5.6-luna-pro evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_a84bda106ab3974d69df8143a7ecfd41ae25fa3af4034c9cda5232a494ed381d74%Sconosciuto1
49nvidia/nemotron-3-super-120b-a12b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_65206ed835937e4a58e9249a2680b4d2872397bf3abf72b22d2c07dc044ed27174%Sconosciuto1
50qwen/qwen3.6-35b-a3b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_099dfd7875cbd5ab4e85221be7656ef51863c17bb94c9af651a328e19a70a2e874%Sconosciuto1
51deepseek/deepseek-v4-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_6bd0c5eb925bb5e63bef71def75efe0301ca52099eca2e0bbcf7981e2aceb51c73%Sconosciuto1
52openai/gpt-6-luna evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_4dbbd2f0ab0d84b7c248fbce5e80ca7551aad370b42a727b4f69166c2026644073%Sconosciuto1
53openai/gpt-5.4-mini evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_4b66200184124eb485b499c0dd56642e2430b644289885afea04aa641c22cda772%Sconosciuto1
54openai/gpt-5.6-luna evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_5cde82caefa8a815fe6f4beb950c65f9fae31efceeda19d1f631601ccdd4b6d372%Sconosciuto1
55deepseek/deepseek-v4-flash-0731 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_e28e0dd0460ebad70a0f394de212558dcc7fb3d62e127c067a94eea4d50e718461%Sconosciuto1
56z-ai/glm-4.7-flash evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_79cff7558daca1c8edbb265f6a708011a390a18f0d07f3295f2944d16c65273159%Sconosciuto1
57openai/gpt-5.1-codex evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_843a8c73b4ee4d1e2dca77b266a2d0a8c20288a24b00117aa916efad91b9562259%Sconosciuto1
58openai/gpt-5.1-codex-mini evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_7791eec96f3a1571763eb0485f92474fc5c2326ef8138f86f49d13bdc8b9a5de57%Sconosciuto1
59z-ai/glm-5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_d92561080dcb2ec18134f028e6d589c34c5ff86933ce31cd0baef9e79281047050%Sconosciuto1
60nvidia/nemotron-3.5-lightning evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_0d55bafe2f486b6d40691b51a77137b641a3498419487b57ed2952d7d09a348349%Sconosciuto1
61openai/gpt-5.1-codex-max evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_lowconfig_7af97febe265676b35c339b4b079bd74568d699f56190ed97a84580bfe2c2ffb47%Sconosciuto1
62openai/gpt-5.4-nano evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_0ab43c9a147fa575ea2d61319dfac2869684ccb48595d6ba5ef34cca75dc928640%Sconosciuto1
63anthropic/claude-sonnet-5 evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_962021295d532bdfe74c91e367fbb71a3bdb956d6247cd8302ccca6d864db38839%Sconosciuto1
64qwen/qwen3.5-35b-a3b evalrank-content-safety:946af6ce1f3654dc9cdffcd310e5dcbf13b451cfc96766af8572debd3783a687:reasoning_offconfig_8d6e5890ab2b84a55f734f9f18f5b5329413ddfa75155e2ecb7b4bd460f6dcf70%Sconosciuto1
Confronta le configurazioni di questo gruppo
Non ci sono abbastanza evidenze indipendenti per un gruppo di testa
  • La calibrazione non è stata convalidata
  • La copertura indipendente delle famiglie di benchmark è sotto la soglia richiesta
Torna a tutti i tipi di lavoro