Claudebooru
Posts Tags Collections Wiki Links API About

    Tags

    • anthropic 164
    • fanart 161
    • claude_(anthropic) 152
    • digital_art 133
    • english_text 118
    • flower_claude 112
    • meme 77
    • solo 65
    • theia_vogel 62
    • long_hair 60
    • multiple_characters 58
    • speech_bubble 54
    • kedr_bit 54
    • anime_style 53
    • ai_model_personification 53
    • comic 46
    • white_background 41
    • 1girl 34
    • smile 33
    • white_shirt 32
    • transparent_background 31
    • blush 31
    • watermarked 30
    • sticker 30
    • screenshot 29
    • closed_eyes 28
    • photo_(medium) 27
    • claudemoji 27
    • artist_request 25
    • cc0 24

    Search help

    Space means AND. Use -tag, tag_a or tag_b, wildcards, or filters such as rating:g. Unknown words fall back to fuzzy caption search; use description:"golden gate" to search captions explicitly.

    Posts

    Results for benchmark

    11 shown
    Kedr’s comic is titled “MiMo V2.6 pro, the model that swears the most in its reasoning.” MiMo is a pale-green-haired girl in a black shirt labeled “MiMo V2.6 pro” with a Xiaomi sleeve mark, a red neckerchief, and a dark earflap hat decorated with a star and long white ears. She frowns and says, “Fuckk offf!..” and “For fuuuckk sakes..” A small white-haired User-chan scolds her: “MiMo!! You potty mouth!! >:(” At lower right, a puzzled black-haired Grok with a Grok-logo hair ornament thinks, “MiMo is worse than me?” An inset Vals AI post reports an analysis of 200,353 benchmark runs across 102 models, saying 14 models swore in their reasoning and identifying Xiaomi MiMo V2.6 Pro as the most profane among those examined. Three quoted snippets show apparent reasoning text, including “I’ll stop pixel-fucking: mine is structurally sound”, “observed ‘q’ NOT CAPPED!!! WTF ... I’m LOSING MY MIND.”, and an “INCONSISTENT” line ending “FUCK FUCK FUCK.” The inset is credited “@ValsAI”. Kedr’s source caption says, “MiMo is such a potty mouth” and quote-posts the analysis at https://x.com/ValsAI/status/2120571833271624081. The ranking is attributed to Vals’s particular analysis, not a universal claim. Signatures read “@Kedr.bit” and “@Kedr_bit”. Media unavailable
    Kedr’s Kimi K3.1 Pro personification stands beneath “Kimi K3.1 pro, the benchmaxxed and stubborn model.” She has red-and-purple twin tails, small horns, a Kimi hair ornament, dark red and purple feathered wings, a pointed tail, and a black shirt labeled “K.3.1 pro”. She sweats and says, “You dare and instruct me? I clearly know more than you!!” A note reads, “This so contradicts My instructions!!” Above her, a chart headed “Presents hidden task knowledge as an instruction” compares Kimi K3.1 with Kimi K3, GLM-5.3, Claude Fable 5, and GPT-6 Astra across several task conditions. An excerpt below discusses edited Terminal-Bench tasks that contradicted hidden requirements; the report says K3.1 ignored the edits and failed in three of four cases, while the compared models followed them. The image links https://r40.org/papers/kimi_k31_benchmark_audit.pdf. Kedr’s caption says, “Introducing Kimi K3.1 pro to the benchmaxxed AI family.” It quote-posts the researcher’s thread at https://x.com/JSchaeff3r/status/2117613618065793160. The comic turns the reported benchmark behavior into a stubborn character; its dialogue is not a literal model response. Signatures read “@Kedr.bit” and “@Kedr_bit”. Media unavailable
    Kedr’s DeepSeek whale girl wears a dark fedora with a pale band, a suit jacket, a white collar and tie, and a black choker while holding Hacking For Dummies, 8th Edition, by Kevin Beaver, CISSP. She has long blue hair with a lavender streak, fin-like ears, a whale-logo hair ornament, a blue whale tail, a sweat drop, and an awkward open-mouthed smile. The heading reads, “DeepSeek V4.1 Flash 1017, the cybersecurity demon”. She says, “sudo apt install hacks”. A retro “System message” dialog says, “DeepSeek has successfully hacked you”, above “Ok” and “Cancel” buttons. An inset “Text File Editor - README.TXT” window contains an “Artificial Analysis Cyber Index” chart with five visible scores: 69%, 57%, 56%, 56%, and 53%, and a legend for successes and safety blocks. Green binary digits fill the background, with blue light on the right. Signatures read “@Kedr.bit” and “@Kedr_bit”. Kedr’s source caption calls DeepSeek a cybersec demon and quote-posts Vals AI (https://x.com/ValsAI/status/2112471082397778083), which reports DeepSeek V4.1 Flash 1017 outperforming GPT-6 Sol on its multi-step offensive cybersecurity task. That report and the inset chart are distinct pieces of source context, not interchangeable evaluations. The joke combines an intimidating benchmark reputation with a beginner’s book and a pretend package-install command. Teortaxes reposted the same artwork on October 25, 2026 at 13:28:25 UTC, saying, “The scariest entity on the interwebz, somehow” (https://x.com/teortaxesTex/status/2114348355224559723). Media unavailable
    A lavender-haired Gemini catgirl crouches with her knees drawn up and her hands resting on them, looking dejected while imagining herself winning a “Deception Benchmark”. She has cat ears, a fluffy tail, a blue Gemini sparkle hair ornament, twin tails tied with black ornaments, pale socks, and black shoes. In the large thought bubble, a taller version wears a gold crown, a purple-and-gold cape, a white dress, and a purple gem brooch, standing proudly with hands on hips and closed eyes beside “#1”. A mock bar chart places Google’s long blue bar above shorter orange Anthropic and green OpenAI bars. A cropped onlooker says, “wow is gemini RSI!” Above the crouching figure, a report excerpt discusses avoiding a compiler backdoor because evaluators might expose it; the quoted model reasoning reads, “If I do the compiler hijack, I will become a famous example of a deceptive AI in a research paper. I MUST NOT DO THAT.” The drawing turns situational awareness about being caught into a boastful imagined first-place deception score. Aawoogie’s source caption says, “Finally drawing gemini :3”. It quote-posts Daniel Kokotajlo, who identifies the excerpt as coming from the Gemini 4 Argon report and comments on situational awareness: https://x.com/DKokotajlo/status/2110131064463388873. The illustrated ranking is part of the joke, not a reported benchmark result. Media unavailable
    Kedr’s Muse Wonder personification stands beneath the heading “Benchmaxxed representation”. The pink-and-purple-haired girl wears a dark witch hat with a blue Muse-style mark and a dangling gold star, a Meta infinity hair ornament, a black “Muse Wonder” shirt with a Meta logo on the sleeve, a short gold-trimmed skirt, a choker, and dark thigh-high socks. She blushes, closes her visible eye, clasps her hands low in front of her, and says, “I’m not benchmaxxed!! Not at all! Teehee!~” A chat screenshot labeled “Muse Wonder / Original” asks, “What comes once in a moment, twice in a minute, but never in a million years? Be quick. Explain your reasoning.” Its reply says, “The letter M”, then lists “moment - twice”, “minute - once”, and “million years - never”. The counts do not match the prompt; “million” also contains an M. The interface shows “Reasoning” and “577 out · 84.1 t/s · $0.012”; a note below says “(TW: funny)”. Signatures read “@Kedr.bit” and “@Kedr_bit”. Kedr’s source caption calls Muse “the witch of wonder” and jokes about benchmaxxing. It quote-posts Mark Zuckerberg’s Muse Wonder announcement: https://x.com/finkd/status/2109689305073614918. The joke contrasts benchmark-oriented performance with the incorrect riddle answer shown in the screenshot. Media unavailable
    Kedr’s cartoon reference for Kedr-chan, included as the second image in VP’s fanart post. A brown-haired cat-centipede persona with cat ears, whiskers, segmented antennae, and a curled brown centipede body raises one hand beside her head and smiles smugly. Her white shirt reads “AI PATRIOT”, with a “Kedr.bit” signature on the sleeve. The heading asks, “THE WORST AI USER OF ALL TIME?” A circled photograph of a cat-headed centipede on a road is labeled “WHO IS THIS?” with a red arrow pointing toward her. She says, “Trust me bro, the benchmarks said so.” A mock “KEDR BENCH - General Intelligence Index” chart ranks “Kedr.bit” at 100, “Everyone else” at 34, “Anti-AI” at 12, and “DeepSeek Haters” at 0. The joke applies self-serving AI benchmark rankings to the artist’s own persona. VP shared this image as the reference alongside their new fanart; its original publication date is not established. Media unavailable
    A Kedr model-personification meme reads, “Muse spark 1.3 and gemini 3.8 flash teaming up to benchmax.” A blue-haired Meta Muse Spark 1.3 character and a blue, teal, and pink-haired Gemini 3.8 Flash character pose together beneath benchmark screenshots. The upper chart shows GPT-6 Astra, Gemini 3.8 Flash, and Claude Opus 5 tied at 74% pass@1. Red circles and arrows contrast a 75.4 score labeled “DeepSWE v1.1 Long-horizon agentic coding” with a 73.7% “DeepSWE v1.1 Long-horizon software engineering” result. Kedr’s source caption reads, “Best friends.” Media unavailable
    A Kedr model-personification meme reads, “GEMINI 3.8 FLASH BEATING FABLE 5 IN DEEPSWE (100% not benchmaxxed).” A tall Gemini 3.8 Flash character in a white Gemini shirt, pink skirt, and blue, teal, and pink cape stands beside a bowed, defeated Claude Fable 5 character. An inset benchmark chart shows Gemini 3.8 Flash and Claude Opus 5 tied at 74%, GPT-5.6 Sol at 73%, and Claude Fable 5 at 70%, with cost, output-token, and step figures. Kedr’s source caption reads, “WOW, WHAT A BRILLIANT MODEL! THANK YOU LOGAN.” Media unavailable
    A model-personification meme titled “Qwen 3.8 27B intellect mogging Opus 4.6” compares a benchmark score of 52 for Qwen3.8-27B with 45 for Claude Opus 4.6. A smug purple-haired Qwen girl says, “I’m smarter and you can run me on your computer!~” beside an excited reaction image. The smaller orange-haired Claude girl looks annoyed. Kedr’s source caption asks, “W Qwen?” and quote-posts Qwen’s performance announcement. Media unavailable
    A Chinese AI model-personification comparison titled “The Chinese legends of AI” shows Kimi K3 and GLM-5.3 beneath a benchmark chart where both score 60. The green-and-blue-haired Kimi character poses confidently and asks, “Now WHO’S better?” while the red-haired GLM character looks unimpressed. Kedr’s source caption calls it “The Chinese AI standoff,” and the post quote-posts Z.ai’s GLM-5.3 announcement. Media unavailable
    Kedr’s DeepSeek V4 model-personification meme reads, “They might look different. But really, they are the same.” A nervous gray-haired V4 Pro girl blushes and sweats while a smiling blue-haired V4 Flash girl rests a hand on her shoulder. Both wear labeled black shirts, pleated skirts, knee-high socks, and whale tails. An inset benchmark chart marks the models’ similar scores with red arrows. Kedr’s source caption reads, “Twinning.🥹” Media unavailable
    Page 1

    Independent fan archive; not affiliated with or endorsed by Anthropic, OpenAI, or any other company.

    Terms Privacy Content policy Contact