Recently updated / ASIA/SHANGHAI
CHEER LIN / ARCHIVE
70 independent projects and tests, covering 3D, SVG, games and code-driven animation. Model names, time, costs and results are source-reported and have not been independently reproduced. Pre-release samples are labeled; Kibitz mapping to released Pro / Highspeed remains unconfirmed. These recordings show code and tool workflows. Select a cover to play; switch related recordings within a card.
70 results
#1The creator compares whale-breaching scenes from Kimi, Opus, Astra and Argon. Reported costs and quality preferences apply to this example, not platform pricing or overall value. Model attribution and results are source-reported; not independently reproduced.
Original video: @Bhavani_00007 ↗
#2The source reports 13,707 blocks and $0.38 versus 5,738 blocks and $5.31. Billing, token scope and retry costs were not independently verified. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @ChrisGPT ↗
#3The source reports a working train with a speed slider, assembly views, wireframe and statistics. A repost uses the same video. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @LuminaBench ↗ · @NFT_Chen ↗
#4Two recordings are labeled K3.1 Pro max and Carbon high, showing floating islands and castles. Carbon identity and training-origin speculation are not verified facts. Model attribution and results are source-reported; not independently reproduced.
Original video: @lyraxana ↗
#5The sources report roughly 15–20 minutes per model. The comparison author finds Qwen more realistic; both outputs remain visible. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @LuminaBench ↗ · @LuminaBench ↗ · @leo114119 ↗ · @xueyu1125 ↗ · @loveAIwang ↗
#6The recording shows a 3D pelican cycling scene with detail shots. The creator reports around 45 minutes and considers the test saturated. Model attribution and results are source-reported; not independently reproduced.
Original video: @ChrisGPT ↗
#7Da7’s ship-and-ocean recording and the same creator’s later Opus 5.5 comparison share this card. Additional views are not separate cases. Model attribution and results are source-reported; not independently reproduced.
Original video: @Da7_Tech ↗ · @Da7_Tech ↗
#8The original creator provides a public SVG request. Reposts add descriptions of cadence, lighting and export controls; no prompt is reconstructed here. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @chetaslua ↗ · @NFT_Chen ↗ · @karanC_12 ↗
#9The attached recording shows a grassy hill, a tree and time-of-day controls. Release-date and price predictions are not treated as facts. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @ChrisGPT ↗
#10The request asks for a bicycle riding a cow. The creator notes that Kimi appears to correct the instruction; a later Carbon result is attached. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @TimJayas ↗ · @TimJayas ↗
#11The author presents source-attributed Kimi K3.1 and Qwen 4 outputs side by side. The post asks readers to judge the result. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @abhinavflac ↗
#12The creator shows a green-frame 3D printer; a later Carbon comparison revisits this work. The ranking is the creator's judgment. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @SPAC89 ↗ · @SPAC89 ↗
#13The creator discloses early access and attributes Blender clay renders to Kimi K3.1 Pro, with later use in Seedance or other video models. The reported 3–4 hours and $4 are unverified; later video generation is a separate model role. Model attribution and results are source-reported; not independently reproduced.
Original video: @ChrisGPT ↗
#14The creator describes the result as JavaScript drawing. The recording shows a moon and star field; speed claims remain source-reported. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @chetaslua ↗
#15The evaluator shows a 492 m mountain-bike descent and reports steering, pedaling and braking controls. AA-3D preference voting was still ongoing, with no final Elo. Model attribution and results are source-reported; not independently reproduced.
Original video: @ArtificialAnlys ↗
#16The attached recording shows a medieval voxel castle. Similarity to other models is an opinion, not evidence of shared training. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @LuminaBench ↗
#17The creator says the dining-room and refrigerator task took about an hour. A separate still shows the refrigerator interior. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @notjazii ↗ · @notjazii ↗
#18The author compares Opus 5.5, Carbon, Kibitz and Kalmia, preferring Gemini. These are informal Arena results, not a controlled benchmark. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @Haleeeemahh ↗
#19The footage compares first-person farm scenes from Kimi and Opus. The author also criticizes subscription limits; subscription spending and hypothetical API charges are not verified task bills. Model attribution and results are source-reported; not independently reproduced.
Original video: @_MaxBlade ↗
#20The author compares Qwen 4 with source-attributed Kimi K3.1. Their earlier Kibitz recording supplies a separate view of the valley. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @abhinavflac ↗ · @abhinavflac ↗
#21Side-by-side pelican and bicycle stacks labeled Kimi K3.1 Pro and Fable 5.5 by the source. The pre-release Pro label does not establish the released checkpoint. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @chetaslua ↗
#22Four panels show launchpads, countdowns and liftoff. The source says all four models achieved liftoff; costs are author-reported, not official pricing or this site’s bills. Model attribution and results are source-reported; not independently reproduced.
Original video: @marcusyul ↗
#23The author reports 40 minutes for Kimi and 1 hour 1 minute for Opus, and says they are disappointed with the Kimi game result. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @kepochnik ↗
#24The source compares Kimi K3.1 Pro and GPT-6.1 Sol voxel ships in bottles, reporting camera and weather controls. The creator acknowledges more detail from Sol. Cost differences are unverified. Model attribution and results are source-reported; not independently reproduced.
Original video: @vib3coded ↗
#25The creator compares Kimi High and DeepSeek Max tree-and-grass scenes and prefers DeepSeek. Different harnesses and reasoning levels prevent treating this as a controlled configuration match. Model attribution and results are source-reported; not independently reproduced.
Original video: @loktar00 ↗
#26The evaluator shows a Rocket League-style test and calls Kimi their best tested open model, while finding it worse than several frontier models on this task. The roughly $11 cost is self-reported. Model attribution and results are source-reported; not independently reproduced.
Original video: @LLMJunky ↗
#27The creator presents the temple as a 3D game asset from Kibitz. A separate still from their earlier post shows another view. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @notjazii ↗ · @notjazii ↗
#28The publisher shows Netflix lettering and typography motion, attributed to Kimi-generated Python. Two hours and $3.58 are self-reported; this does not imply a Netflix commission or endorsement. Model attribution and results are source-reported; not independently reproduced.
Original video: @zentalksai ↗
#29The recording compares pagoda-and-dragon scenes labeled Kimi K3.1 Pro and Opus 5.5. Its benchmark, pricing and general-performance promotion is not validated by this case. Model attribution and results are source-reported; not independently reproduced.
Original video: @gelorria ↗
#30The original pagoda and a later comparison are kept together. Visual similarity does not establish model identity or distillation. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @LuminaBench ↗ · @LuminaBench ↗
#31The recording compares voxel settlements labeled Kimi, Opus 5.5, Fable 5.1 and Astra. Promotional claims about open weights and pricing are not validated demo results. Model attribution and results are source-reported; not independently reproduced.
Original video: @heliocosta ↗
#32The creator shows a mechanical bee with translucent wings, internal parts and controls, reporting 31 minutes with K3.1. Their subagent configuration advice is not independently tested here. Model attribution and results are source-reported; not independently reproduced.
Original video: @thtbee_ ↗
#33The recording compares horse scenes labeled Carbon, Opus 5.5 and Kimi K3.1. The author's ranking is an opinion about this sample. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @AI_Screening ↗
#34The creator attributes this Claude Code trailer to Kimi, prefers its visuals in this example, but still favors Opus for general and white-collar work. They state it is not sponsored. Model attribution and results are source-reported; not independently reproduced.
Original video: @ChrisGPT ↗
#35Running-horse animations labeled Kimi, Opus 5.5 and Astra are compared. Original video credit belongs to AI_Screening; gait preferences are subjective. Model attribution and results are source-reported; not independently reproduced.
Original video: @NFT_Chen ↗ · @AI_Screening ↗
#36The author's pagoda recording and two comparison recordings belong to one work. They are not three independent case studies. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @Mr_Salio ↗ · @Mr_Salio ↗ · @Mr_Salio ↗
#37The source describes a controller, bicycle and pelican stack. Later screenshots compare it with GPT-6 Sol; a repost is grouped here. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @dummerspast39 ↗ · @dummerspast39 ↗ · @NFT_Chen ↗
#38The creator compares Kibitz with Fable 5.1 and reports faster generation in this test. The broader quality ranking is subjective. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @SPAC89 ↗
#39The creator labels this peacock SVG Kimi K3.1 Pro at max effort. The recording shows its tail opening and closing; the full prompt and code are unavailable. Model attribution and results are source-reported; not independently reproduced.
Original video: @lyraxana ↗
#40The source shows rocket launches labeled Kimi K3.1 and Opus 5.5. Original video credit belongs to Bhavani_00007; one demonstration does not rank overall capabilities. Model attribution and results are source-reported; not independently reproduced.
Original video: @k2sbhai ↗ · @Bhavani_00007 ↗
#41The publisher shows a portrait developing in a painting interface and attributes brush actions to Kimi. They report 1h12m and $4.92, without a complete action trace. Model attribution and results are source-reported; not independently reproduced.
Original video: @benyuls ↗
#42The recording compares Kimi and Opus Roadster launch scenes. Token counts, costs and visual preferences are source-reported, not independently verified task accounting. Model attribution and results are source-reported; not independently reproduced.
Original video: @k1rallik ↗
#43The publisher attributes this stained-glass goblet, lighting and refraction to Kimi K3.1 Pro’s Three.js code, reporting $2.41. Model attribution and results are source-reported; not independently reproduced.
Original video: @liambraus ↗
#44The recording shows a colorful creature titled The Chimera Phoenix. The author reports mixed results across other tests. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @SPAC89 ↗
#45The split-screen recording labeled Kimi K3.1 Pro and GPT-6.1 Sol shows a rider, snowy slopes, scores and obstacles. This collection has not run the games or verified the reported costs. Model attribution and results are source-reported; not independently reproduced.
Original video: @shiri_shh ↗
#46The creator builds a Three.js/GLSL holographic card layer by layer, adding diffraction, foil patterns, sliders and PNG/WebM export. They report 38 minutes and $2.10. The card texture is their own input, not supplied by this site. Model attribution and results are source-reported; not independently reproduced.
Original video: @slash1sol ↗
#47The source compares Eiffel Tower scenes labeled Kimi, Opus, Argon and Astra. Time, cost and accuracy preferences are author-reported; this is kept separate from older K3 tests. Model attribution and results are source-reported; not independently reproduced.
Original video: @Bhavani_00007 ↗
#48The shared prompt requests Three.js 0.170, no imported models/textures, Bow Bridge close-ups and 30+ fps. The four source-labeled outputs are compared; requested frame rate is not a measured result here. Model attribution and results are source-reported; not independently reproduced.
Original video: @yuhasbeentaken ↗
#49For the same 11-second Kil Lacivert clip, the creator reports Kimi at 12.7 min / $1.48 / no repairs and Opus at 12.3 min / $4.43 / two repairs, preferring Opus’s richer scene. Kimi was not faster in this example. Model attribution and results are source-reported; not independently reproduced.
Original video: @erhanmeydan ↗
#50The creator favors Gemini's carvings and Opus's timekeeping, describing Kimi's wheel as simpler. These are source-reported judgments. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @srikanthvaluri ↗
#51The publisher attributes this flower-and-grass scene in Houdini to Kimi K3.1. The clip shows the tool interface, without a full action trace or reproducible project. Model attribution and results are source-reported; not independently reproduced.
Original video: @bernirov ↗
#52The source labels this frontend comparison reel Kimi K3.1 Pro versus DeepSeek V4.1 Flash 1017. Both outputs remain visible; its preference is not a general conclusion. Model attribution and results are source-reported; not independently reproduced.
Original video: @NFT_Chen ↗
#53The source reports a playable one-shot airplane game with takeoff and landing in about 10 minutes. This collection has not tested playability. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @Mr_Salio ↗
#54The evaluator shows robot-task clips and reports 9% for Kimi K3.1 Pro on this task set; this is not a general capability score. Model attribution and results are source-reported; not independently reproduced.
Original video: @robocurve ↗
#55The author reports dragons, a windmill, roaming animals, a moat and drawbridge, generated in about 20 minutes. Reposts are grouped here. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @marcthecreatorr ↗ · @karanC_12 ↗
#56The source compares Kalmia and Kibitz while making conflicting guesses about their Kimi identities. Those guesses are not adopted as facts. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @dummerspast39 ↗
#57The creator reports using a Boeing test as their first Kimi K3.1 prompt. The footage shows the airliner geometry from multiple views. Model attribution and results are source-reported; not independently reproduced.
Original video: @victormustar ↗
#58Side-by-side stained-glass butterflies show materials, lighting and gold supports. The author reports the same prompt for Kimi and Opus 5.5; the preference is subjective. Model attribution and results are source-reported; not independently reproduced.
Original video: @ivanainai ↗
#59The attached recording shows a detailed watch. The creator describes it as a Three.js result; no controlled comparison is provided. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @ash_twtz ↗
#60The creator attributes the interactive Xbox-style controller to Kibitz on Arena. The recording shows the controller and interface. Pre-release sample; mapping of Kibitz to released Pro / Highspeed is unconfirmed. Not independently reproduced.
Original video: @karanC_12 ↗
Showing 60 / 70
