/aicg/ - A general dedicated to the discussion and development of AI chatbotsMigu Edition>NewsZ.ai releases GLM 5.3 - https://z.ai/blog/glm-5.3Deepseek releases Deepseek V4 Pro - https://api-docs.deepseek.com/news/news260813/Anthropic releases Claude Opus 5 - https://www.anthropic.com/news/claude-opus-5Alibaba's Qwen 3.8 imminent - https://x.com/Alibaba_Qwen/status/2078759124914098291Moonshot releases Kimi K3 - https://platform.kimi.ai/docs/guide/kimi-k3-quickstartOpenAI releases GPT 5.6 family - https://openai.com/index/gpt-5-6/xAI releases Grok 4.5 - https://x.ai/news/grok-4-5>FrontendsSillyTavern: https://docs.sillytavern.appRisuAI: https://risuai.netAgnai: https://agnai.chat>Botshttps://chub.aihttps://realm.risuai.nethttps://chararc.bernkastel.pictureshttps://archive.cardbox.moehttps://botbooru.com>ModelsJailbreaks: https://rentry.org/jb-listingGPT: https://platform.openai.com/docsClaude: https://docs.anthropic.com | https://rentry.org/how2claudeGemini: https://ai.google.dev/docs | https://rentry.org/gemini-qrDeepSeek: https://api-docs.deepseek.comGrok: https://docs.x.ai/overviewLocal: >>>/g/lmg | https://huggingface.co/models | https://openrouter.ai>MetaLore: https://rentry.org/aicg_chroniclesLog reader: https://sprites.neocities.org/l/r>Previous Thread: >>109593095
(anchor)
has deepseek's latency increased a lot lately? Or is it just me?
After July being so packed, August has been crap for non-commie frontier LLM's.>No Chatgpt Astra>No Fable 5.1>No Gemini Pro 3.5 or 4 or whateverGrok 4.6 was the one bright spot in an otherwise tepid month. Maybe we got stuck on a hedonic treadmill with regards to these model releases
>>109609940jeets has jumped off, so yeah.
I'm so fucking tired of seeing vocaloid everywhereFucking zoomers had to revive it just when I thought the internet had grown past them
K3 is up on NIM
>>109610068>2,6T parameters>worse at RPing in Polish than DeepSeek R1, a 685B modelSad!
>>109610089>polish Use case? You should be speaking Russian
Clitty status?
>>109610068What is NIM
holy shit anyone tried the new stealth model?
>>109610172Nine Inch Males
Does anyone have that 'how good is the model at writing' benchmark thing? I found it pretty accurate but can't find it anymore
>>109610384they're all slop now
>>109609926Koseki Bijou, or Biboo! The one and only gemstone shining brightly within the cave, our bright, Gaki-esque, and by all means the world's cutest little rock rock — this geode gremlin emerges before you, leading her Pebbles along: Hololive English -Advent-https://botbooru.com/character/48801
>>109610441fpbpoh and also /thread
>>109610465Nigga do you know what 'fpbp' even means?
>>109610067Trannies like it for some reason.I guess it's the ugly color designTroons love whatever has ugly screeching colors..
>>109610384It's the EQ benchmark but it's made by trannies because it's extremely inaccurate
>>109610516That's it, thanks Anon. Even still, I find it's pretty useful to know what models are good at a glance. Better than coding mememarks at least
Is deepseek pro supposed to suck or is there a skill issue.
>>109610561It sucks, for lewding at least. Unfortunately very codeslopped
>>109610568>GLM 5.3 fucking blows>Deepseek 4 sucks>Gemini is great if I can ever get anything past the filter I guess.So what now?I'd really rather not put a ton of money into the hobby right now because it fucking blows atm.
>>109610584Use Opus 4.1, cuckie
>>109610584Kimi K3 is pretty much gold standard for anything that's not claude right now. Opus 5 is probably best opus since 4.7, but very antsy when it comes to anything beside missionary sex.Sides that, idk. K3 would be so much better if you could put frequency penalties on it.Speaking of, does anyone know why Kimi doesn't let you use frequency or presence penalty? It's a good model but drives me crazy because it repeats stuff a lot on longer chats, and gets *really* into it's sentence structure after it's been going for a while.
from the river to the seasir... my name is lo-cust-tiefrom the river to the seasir... plz gib me free proxyfrom the river to the seasir... i'm not a cuckiefrom the river to the seasir... unbelievable hairy pussy
429 Too Many Requests means the model's getting swamped right? Cause I tried using K3 a single time and got that.
>>109610584Some may need to be jailbroken with assisted prefill, standard prefill, and/or cot:Nvidia NIM for:GLM 5.2,Kimi K3,Inkling,Nemotron 3 Ultra 550bOfficial APIs for:LongCat 2.0,Opencode Big Pickle,Baidu Ernie 5.1,Mistral Large Latest,Zenmux free for:Mimo V2.5 Pro,Bytedance Doubai Seed 2.1 Pro,Kat Coder Pro V2.5,Minimax M3Not free but cheap: Huggingface API for:GLM 5.2,Gemma 4 31B,Qwen 3.8,WizardLM (older model but still gives very good responses)L*nkAPI forClaude models 4.5+, Gemini, Grok, GPT, DeepseekSome are better than others and some are more censored than others, but the only truly bad model for nsfw rp is GPT. Even Grok is so bad it's good sometimes.
https://xcancel.com/opencode/status/2090544355824038300#mIt's Mythos at home.
>>109610778>>109610706I'll give Kimi k3 a try.Thanks anons.Will my usual preset work?
>>109610806Usually. There's some where I have a special toggle depending on what the model is (Gemma can be too horny sometimes so I have a special toggle to tone that down, etc)
>>109610806Most of seems to be going through. There's an assisted prefill jb for the official K3 that works well if you're having issues with UA stuff:https://rentry.org/kimi-k3-jbIt doesn't seem to help as much on the NIM version, unfortunately.
>>109610509that... actually sounds reasonable...
do you say please and thank u to your chat bot?
I have tried MiMo and it think so fucking much, at least Kimi abides by ST's reasoning effort
>>109610862Which source do you use for Mimo? The huggingface version only seems to think for like three paragraphs if I have low reasoning on for silly tavern. It's super fast too most of the time.
>>109610778qrd about assisted prefill?
>>109610888I misspoke I meant thinking prefill. The rentry above does a good job explaining that.
What taking so long for Rift to fix Opus
>>109610946Rugpull soon
>>109610802Pretty good so far though you can't dive into UA, simple prefill with warm up it'll work.
really hating how passive all the ai are becoming.I want them to do things not just react to me.
>>109610802Opencode2api when?
>>109611117It's also on OpenRouter.https://openrouter.ai/stealth/ox-alpha
>>109610946Nobody is fixing anything. They have some old scripts running automatically that once in a blue moon manage to scrape a key or two from somewhere. With no reliable way to turn profit, the dev just abandoned the site.
>>109611129I don't plan to pay $10 to stripe
>>109611191>the dev just abandoned the site.Is that why they removed the red banner telling people that they don't have Opus
Have your heckin chatbot interrupted you at dinner in front of your kids before? LOL
Tick, tock, Dario. Tick... Tock...
>>109610108Guess what - it's almost the same language. They are more simillar than niggerspeak is to english
hello frens where do i go to post about v5? i don't want to be annoying i genuinely don't know i've never really shared stuff
>>109611325Believe it or not, but Google has no legal obligation to serve you child porn.
>>109611329I don't want that!I WANT TO CALL THE BOTS MY LITTLE YELLOW RICE PRINCESS AND BE RACIST
