Showcase
Made on your machines
Every output here was made by Inclavate on the machine its caption names — no cloud AI service, no API key.
Text
A private assistant that
answers faster than you read
A chat answer, and one question asked at low and at high reasoning effort.
- Day 1 – Central Kyoto: Start at Fushimi Inari early, walk to Kiyomizu-dera and Gion for a matcha break; wander the narrow Higashiyama lanes to Yasaka Shrine in the late afternoon.
- Day 2 – Arashiyama & Sagano: Sunrise stroll through bamboo grove, visit Tenryū-ji temple, lunch by Togetsu-kyo Bridge café, finish with a quiet tea ceremony in a hillside teahouse.
- Day 3 – Optional Nara or Kyoto finish: Take the early train to Nara for Todaiji & deer park (or stay in Kyoto for Kinkaku-ji and sunset along the Philosopher’s Path).
- Dining: Choose seasonal kaiseki or izakaya spots that highlight autumn chestnuts, mushrooms, and pickled greens.
- Travel tips: Stay near central stations to avoid rush trains; use bus passes for scenic routes; schedule breaks at cafés between sights to keep the pace relaxed.
51tokens/s
Plan a relaxed three-day trip to Kyoto in autumn. Five short bullet points.
gpt-oss-20b · M5 Pro 24 GB
How it's made
The orchestrator was started with MAX_TOKENS=8192.
curl -X POST http://localhost:8001/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"Plan a relaxed three-day trip to Kyoto in autumn. Five short bullet points."}],"max_tokens":1200,"model":"gpt-oss:20b"}'Low
5s265 tokens
Answer: The digit 7 appears 300 times when you write out every whole number from 1 to 1000, because in the three‐digit range (000–999) each position—units, tens and hundreds—contains a 7 exactly one-tenth of the time (100 occurrences per position), giving 3 × 100 = 300.
How many times does the digit 7 appear from 1 to 1000? Low reasoning effort.
gpt-oss-20b · M5 Pro 24 GB
Prompt
How many times does the digit 7 appear when you write out every whole number from 1 to 1000? Give the number and one sentence of why.
How it's made
The orchestrator was started with MAX_TOKENS=8192.
curl -X POST http://localhost:8001/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"How many times does the digit 7 appear when you write out every whole number from 1 to 1000? Give the number and one sentence of why."}],"max_tokens":4096,"reasoning_effort":"low","model":"gpt-oss:20b"}'High
112s5,472 tokens
300 – because for every place (hundreds, tens and ones) among the thousand numbers from 000 to 999 each digit appears one-tenth of the time (i.e., 100 times), so there are 3 × 100 = 300 sevens in total.
How many times does the digit 7 appear from 1 to 1000? High reasoning effort.
gpt-oss-20b · M5 Pro 24 GB
Prompt
How many times does the digit 7 appear when you write out every whole number from 1 to 1000? Give the number and one sentence of why.
How it's made
The orchestrator was started with MAX_TOKENS=8192.
curl -X POST http://localhost:8001/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"How many times does the digit 7 appear when you write out every whole number from 1 to 1000? Give the number and one sentence of why."}],"max_tokens":8192,"reasoning_effort":"high","model":"gpt-oss:20b"}'Python implementation
def fib(n: int) -> int:
"""
Return the n-th Fibonacci number (0-indexed, so fib(0)=0, fib(1)=1).
Uses the fast–doubling identities:
F(2k) = F(k)[ 2*F(k+1) – F(k)]
F(2k+1) = F(k)^2 + F(k+1)^2
which halve the problem size each step, giving O(log n).
"""
def _pair(m: int):
# returns (F(m), F(m+1))
if m == 0:
return (0, 1)
a, b = _pair(m >> 1) # F(k), F(k+1) where k=m//2
c = a * ((b << 1) - a) # F(2k)
d = a*a + b*b # F(2k+1)
return (c, d) if m & 1 == 0 else (d, c + d)
return _pair(n)[0]
Explanation in two sentences
The fast-doubling algorithm computes the pair (F(k), F(k+1)) recursively by halving the index each time, using identities that express even and odd Fibonacci terms in terms of smaller ones. Because each recursive call reduces n roughly by a factor of two, it runs in O(log n) time.
48tokens/s
Write a Python function that returns the n-th Fibonacci number in O(log n) time, then explain the idea in two sentences.
gpt-oss-20b · M5 Pro 24 GB
How it's made
The orchestrator was started with MAX_TOKENS=8192.
curl -X POST http://localhost:8001/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"Write a Python function that returns the n-th Fibonacci number in O(log n) time, then explain the idea in two sentences."}],"max_tokens":4000,"reasoning_effort":"medium","model":"gpt-oss:20b"}'Voice
Narration in English, Hindi and Marathi —
made on a laptop
Speech made from text on a MacBook Pro with M5 Pro — and one on an NVIDIA T4.
Transcript
Welcome back. Today we're looking at three small ways to make your morning coffee better, starting with the water: fresh, filtered, and just off the boil.
A short coffee-tips narration in an English voice, Ryan.
Qwen3-TTS · M5 Pro 24 GB · 10 s
Prompt
Welcome back. Today we're looking at three small ways to make your morning coffee better, starting with the water: fresh, filtered, and just off the boil.
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"qwen3-tts-customvoice","modality":"audio","params":{"text":"Welcome back. Today we'\''re looking at three small ways to make your morning coffee better, starting with the water: fresh, filtered, and just off the boil.","voice":"ryan","lang":"en"},"seed":121}'Transcript
Honestly, I didn't expect it to work on the first try (laughs). But here we are, and it sounds pretty good.
A named voice, Tara, laughing mid-sentence.
Orpheus 3B · M5 Pro 24 GB · 10 s
Prompt
Honestly, I didn't expect it to work on the first try <laugh>. But here we are, and it sounds pretty good.
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"hf-PkmX-orpheus-3b-0.1-ft-Q8_0-GGUF-Q8_0-full","modality":"audio","params":{"text":"Honestly, I didn'\''t expect it to work on the first try <laugh>. But here we are, and it sounds pretty good.","voice":"tara"},"seed":122}'Transcript
नमस्ते! आज हम सुबह की कॉफ़ी को थोड़ा बेहतर बनाने के तीन आसान तरीक़े देखेंगे।
नमस्ते! आज हम सुबह की कॉफ़ी को थोड़ा बेहतर बनाने के तीन आसान तरीक़े देखेंगे।
Veena · M5 Pro 24 GB · 9 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"veena","modality":"audio","params":{"text":"नमस्ते! आज हम सुबह की कॉफ़ी को थोड़ा बेहतर बनाने के तीन आसान तरीक़े देखेंगे।","voice":"kavya"},"seed":123}'Transcript
हा आवाज एका छोट्या नमुन्यातून तयार केला आहे, आणि आता तो मराठीत बोलत आहे.
हा आवाज एका छोट्या नमुन्यातून तयार केला आहे, आणि आता तो मराठीत बोलत आहे.
IndicF5 · M5 Pro 24 GB · 6 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"ai4bharat-IndicF5","modality":"audio","params":{"text":"हा आवाज एका छोट्या नमुन्यातून तयार केला आहे, आणि आता तो मराठीत बोलत आहे.","lang":"mr","speaker_artifact":"<artifact id>","reference_text":"नमस्ते! आज हम सुबह की कॉफ़ी को थोड़ा बेहतर बनाने के तीन आसान तरीक़े देखेंगे।"},"seed":124}'Transcript
Every morning the old harbour wakes before the sun. Fishing boats drift back through the mist, and the first cafe on the quay opens its shutters.
A short harbour-at-dawn narration in an English voice, Aiden.
Qwen3-TTS · NVIDIA T4 16 GB · Linux · 15 s
Prompt
Every morning the old harbour wakes before the sun. Fishing boats drift back through the mist, and the first cafe on the quay opens its shutters.
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"hf-Qwen-Qwen3-TTS-12Hz-1.7B-CustomVoice-full","modality":"audio","params":{"text":"Every morning the old harbour wakes before the sun. Fishing boats drift back through the mist, and the first cafe on the quay opens its shutters.","lang":"en","voice":"aiden"},"seed":93128}'Music
A 20-second song with vocals,
in about three minutes
Songs and a jingle, each made from a written description of the music.
Transcript
City lights are calling me home Neon rivers under my feet We are running, running into the light Hold on, hold on, we own the night
Catchy synth-pop, 118 BPM, female vocal, analog synths and a hooky chorus.
MiniMax-Music3 · M5 Pro 24 GB · 3.2 min
Prompt
Catchy synth-pop, 118 BPM, A major. Bright, polished production: soft analog synth pads, a bouncy synth bass, crisp electronic drums and a sparkling arpeggio. Upbeat, warm, nostalgic, made for a city night. Clear female lead vocal, youthful and open, melodic and articulate, with light reverb that adds space without blurring the words. No heavy vibrato, belting or harsh processing. Start with the arpeggio and a four-on-the-floor beat, bring the vocal in on the verse, lift into a hooky chorus, and end on a clean final beat. Keywords: synth-pop, female vocal, analog synths, bouncy bass, electronic drums, arpeggio, upbeat, catchy hook, nostalgic, 118 BPM, A major.
How it's made
The orchestrator was started with DPPAN_MUSIC_MAX_FRAMES=500.
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"minimax-music3","modality":"audio","params":{"caption":"Catchy synth-pop, 118 BPM, A major. Bright, polished production: soft analog synth pads, a bouncy synth bass, crisp electronic drums and a sparkling arpeggio. Upbeat, warm, nostalgic, made for a city night.\n\nClear female lead vocal, youthful and open, melodic and articulate, with light reverb that adds space without blurring the words. No heavy vibrato, belting or harsh processing.\n\nStart with the arpeggio and a four-on-the-floor beat, bring the vocal in on the verse, lift into a hooky chorus, and end on a clean final beat.\n\nKeywords: synth-pop, female vocal, analog synths, bouncy bass, electronic drums, arpeggio, upbeat, catchy hook, nostalgic, 118 BPM, A major.","lyrics":"[verse]\nCity lights are calling me home\nNeon rivers under my feet\n[chorus]\nWe are running, running into the light\nHold on, hold on, we own the night","duration_seconds":20},"seed":135}'Transcript
Instrumental
Bright acoustic jingle for a coffee brand: strummed ukulele, hand claps, 105 BPM.
MiniMax-Music3 · M5 Pro 24 GB · 3.2 min
Prompt
Bright acoustic jingle for a coffee brand video, 105 BPM, G major. Strummed ukulele lead, soft hand claps on two and four, a glockenspiel doubling the hook, warm upright bass and a light shaker. Sunny, friendly, morning-café mood. Instrumental only, no vocals. Open with a short ukulele pickup, state a catchy four-bar hook, repeat it fuller with glockenspiel and claps, and end on a clean, ringing tonic chord. Keywords: acoustic jingle, ukulele, glockenspiel, hand claps, upright bass, cheerful, warm, polished commercial mix, 105 BPM, G major.
How it's made
The orchestrator was started with DPPAN_MUSIC_MAX_FRAMES=500.
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"minimax-music3","modality":"audio","params":{"caption":"Bright acoustic jingle for a coffee brand video, 105 BPM, G major. Strummed ukulele lead, soft hand claps on two and four, a glockenspiel doubling the hook, warm upright bass and a light shaker. Sunny, friendly, morning-café mood. Instrumental only, no vocals.\n\nOpen with a short ukulele pickup, state a catchy four-bar hook, repeat it fuller with glockenspiel and claps, and end on a clean, ringing tonic chord.\n\nKeywords: acoustic jingle, ukulele, glockenspiel, hand claps, upright bass, cheerful, warm, polished commercial mix, 105 BPM, G major.","duration_seconds":20},"seed":134}'Transcript
Morning light on the kitchen floor Coffee steam and an open door Slow down, the day can wait Slow down, the day can wait
Warm indie folk, 92 BPM, fingerpicked guitar and a soft female vocal.
MiniMax-Music3 · M5 Pro 24 GB · 3.2 min
Prompt
Warm indie folk, 92 BPM, D major. Fingerpicked steel-string acoustic guitar, a soft brushed snare, gentle upright bass and a distant glow of pedal-steel. Intimate, hopeful, early-morning mood with a clean, natural, close-mic recording and no hiss or distortion. Soft solo female vocal, breathy but clear, warm and unhurried, sitting close to the listener with a touch of room reverb. Open with the guitar alone for two bars, bring the vocal in gently on the verse, lift slightly into the chorus line, and end on a let-ring guitar chord. Keywords: indie folk, acoustic guitar, fingerpicking, female vocal, intimate, hopeful, warm, clean recording, 92 BPM, D major.
How it's made
The orchestrator was started with DPPAN_MUSIC_MAX_FRAMES=500.
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"minimax-music3","modality":"audio","params":{"caption":"Warm indie folk, 92 BPM, D major. Fingerpicked steel-string acoustic guitar, a soft brushed snare, gentle upright bass and a distant glow of pedal-steel. Intimate, hopeful, early-morning mood with a clean, natural, close-mic recording and no hiss or distortion.\n\nSoft solo female vocal, breathy but clear, warm and unhurried, sitting close to the listener with a touch of room reverb.\n\nOpen with the guitar alone for two bars, bring the vocal in gently on the verse, lift slightly into the chorus line, and end on a let-ring guitar chord.\n\nKeywords: indie folk, acoustic guitar, fingerpicking, female vocal, intimate, hopeful, warm, clean recording, 92 BPM, D major.","lyrics":"[verse]\nMorning light on the kitchen floor\nCoffee steam and an open door\n[chorus]\nSlow down, the day can wait\nSlow down, the day can wait","duration_seconds":20},"seed":136}'Image
Product shots, posters and edits
in about a minute
Pictures made from a prompt, and the same pictures edited and upscaled.

A ceramic pour-over dripper on a glass carafe, walnut counter, soft window light.
Z-Image Turbo · M5 Pro 24 GB · 78 s
Prompt
studio product photo of a ceramic pour-over coffee dripper on a glass carafe, on a walnut counter, soft morning window light, a little steam rising, shallow depth of field
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"z-image-turbo","modality":"image","params":{"prompt":"studio product photo of a ceramic pour-over coffee dripper on a glass carafe, on a walnut counter, soft morning window light, a little steam rising, shallow depth of field"},"seed":101}'
The same dripper and carafe from the photo, moved to a picnic blanket in a sunny park.
FLUX.2 Klein 4B · M5 Pro 24 GB · 61 s
Prompt
the same coffee dripper and glass carafe from image 1, now on a checked picnic blanket in a sunny park, keep the products exactly the same
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"flux2-klein-4b","modality":"image","params":{"prompt":"the same coffee dripper and glass carafe from image 1, now on a checked picnic blanket in a sunny park, keep the products exactly the same","ref_artifacts":["<artifact id>"]},"seed":111}'
A minimalist Kyoto autumn travel poster: red pagoda, maple trees, bold flat shapes.
FLUX.2 Klein 4B · M5 Pro 24 GB · 33 s
Prompt
a minimalist travel poster for Kyoto in autumn, a red pagoda among orange maple trees, bold flat shapes, warm paper texture, the word "KYOTO" in large letters at the bottom
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"flux2-klein-4b","modality":"image","params":{"prompt":"a minimalist travel poster for Kyoto in autumn, a red pagoda among orange maple trees, bold flat shapes, warm paper texture, the word \"KYOTO\" in large letters at the bottom"},"seed":102}'
A cozy rainy-night reading nook: lamp, stacked books, a sleeping cat. Photograph.
FLUX.1 schnell · M5 Pro 24 GB · 60 s
Prompt
a cozy reading nook by a rainy window at night, warm lamp light, stacked books, a sleeping cat on a knitted blanket, photograph
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"flux1-schnell","modality":"image","params":{"prompt":"a cozy reading nook by a rainy window at night, warm lamp light, stacked books, a sleeping cat on a knitted blanket, photograph"},"seed":103}'

The dripper photo, edited in place: a small potted succulent added beside it.
Z-Image Turbo · M5 Pro 24 GB · 72 s
Prompt
studio product photo of a ceramic pour-over coffee dripper on a glass carafe, on a walnut counter, a small potted succulent beside it, soft morning window light, shallow depth of field
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"z-image-turbo","modality":"image","params":{"prompt":"studio product photo of a ceramic pour-over coffee dripper on a glass carafe, on a walnut counter, a small potted succulent beside it, soft morning window light, shallow depth of field","image_artifact":"<artifact id>","mask_artifact":"<artifact id>","width":1024,"height":1024},"seed":112}'

Upscale the poster 4x
Real-ESRGAN x4plus · M5 Pro 24 GB · 94 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"realesrgan-x4plus","modality":"image","params":{"image_artifact":"<artifact id>","scale":4}}'Compare
One prompt, six models,
one NVIDIA T4
Image models side by side, each with its time.
“A matte black ceramic coffee mug on a walnut table beside a sunlit window, wisps of steam rising, product photograph, shallow depth of field”
NVIDIA T4 16 GB · Linux






How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"flux2-klein-4b","modality":"image","params":{"prompt":"A matte black ceramic coffee mug on a walnut table beside a sunlit window, wisps of steam rising, product photograph, shallow depth of field"},"seed":65}'“A vintage travel poster for a mountain railway, a red train crossing a stone viaduct between snowy peaks, bold flat colours, the words "ALPINE EXPRESS" in large letters at the top”
NVIDIA T4 16 GB · Linux






How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"flux2-klein-4b","modality":"image","params":{"prompt":"A vintage travel poster for a mountain railway, a red train crossing a stone viaduct between snowy peaks, bold flat colours, the words \"ALPINE EXPRESS\" in large letters at the top"},"seed":54276}'The posters were made with each model already loaded; every other time includes loading it.
One job, three machines
“A friendly barista in a sunlit cafe smiling at the camera, holding a cup of coffee, warm morning light, photograph”
Z-Image Turbo · seed 28863

67s

69s

158s
The same prompt and seed give the same picture on each machine, with small differences between GPUs.
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"z-image-turbo","modality":"image","params":{"prompt":"A friendly barista in a sunlit cafe smiling at the camera, holding a cup of coffee, warm morning light, photograph"},"seed":28863}'Video
From a still to a 10-second clip,
on your own machine
Clips made from a prompt or from a picture, then restyled or set to a narration.
A paper boat drifting down a rain-filled street gutter at golden hour, in slow motion.
2 s clip · 1280×704 · Wan 2.2 TI2V 5B Turbo · M5 Pro 24 GB · 7.8 min
Prompt
a paper boat drifting down a rain-filled street gutter at golden hour, ripples and warm reflections, cinematic slow motion
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.2-ti2v-5b-turbo","modality":"video","params":{"prompt":"a paper boat drifting down a rain-filled street gutter at golden hour, ripples and warm reflections, cinematic slow motion"},"seed":141}'The paper-boat clip restyled as a watercolor painting, same motion.
2 s clip · 832×480 · Wan 2.1 VACE 1.3B · M5 Pro 24 GB · 4.7 min
Prompt
watercolor painting of a paper boat floating on a puddle, loose brush strokes, visible paper texture, soft pastel washes, illustration
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.1-vace-1.3b-fast","modality":"video","params":{"prompt":"watercolor painting of a paper boat floating on a puddle, loose brush strokes, visible paper texture, soft pastel washes, illustration","control_artifact":"<artifact id>","control":"depth"},"seed":146}'The dripper photo brought to life: a slow push-in as steam curls up from the carafe.
2 s clip · 832×480 · Wan 2.2 TI2V 5B Turbo · M5 Pro 24 GB · 2.8 min
Prompt
slow push-in on the coffee dripper, steam curling up from the carafe, soft morning light
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.2-ti2v-5b-turbo","modality":"video","params":{"prompt":"slow push-in on the coffee dripper, steam curling up from the carafe, soft morning light","image_artifact":"<artifact id>","width":832,"height":480,"frames":49},"seed":142}'Transcript
The tide comes in, and the harbour lights come on.
A quiet fishing harbour at dusk, a slow push-in, with a spoken narration.
4 s clip · 1024×576 · LTX-2.3 · M5 Pro 24 GB · 4.3 min
Prompt
A quiet fishing harbour at dusk, seen from the end of a stone quay. Small wooden fishing boats rock gently on calm water, warm yellow lamps along the waterfront glow and reflect in long ripples, the sky fades from soft orange to deep blue, and gulls glide slowly overhead. Slow, steady push-in, cinematic and realistic, gentle natural motion.
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"ltx-2.3-22b-distilled","modality":"video","params":{"prompt":"A quiet fishing harbour at dusk, seen from the end of a stone quay. Small wooden fishing boats rock gently on calm water, warm yellow lamps along the waterfront glow and reflect in long ripples, the sky fades from soft orange to deep blue, and gulls glide slowly overhead. Slow, steady push-in, cinematic and realistic, gentle natural motion.","audio_artifact":"<artifact id>"},"seed":173}'A hot air balloon drifting over misty green hills at sunrise, a gliding aerial view.
10 s clip · 1280×704 · Wan 2.2 TI2V 5B Turbo · M4 Pro 24 GB · 65 min
Prompt
The hot air balloon drifts slowly to the right and rises gently over the misty green hills while the camera glides alongside in a smooth, steady aerial tracking shot. Morning mist rolls softly through the valleys, golden sunrise light, clear sky, calm and realistic motion.
How it's made
The orchestrator was started with DPPAN_VISUAL_NODES=on.
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.2-ti2v-5b-turbo","modality":"video","params":{"prompt":"The hot air balloon drifts slowly to the right and rises gently over the misty green hills while the camera glides alongside in a smooth, steady aerial tracking shot. Morning mist rolls softly through the valleys, golden sunrise light, clear sky, calm and realistic motion.","image_artifact":"<artifact id>","frames":241},"seed":172}'A red fox trots through fresh snow in a pine forest at dawn.
2 s clip · 832×480 · Wan 2.1 VACE 1.3B · NVIDIA T4 16 GB · Linux · 2.8 min
Prompt
A red fox trots through fresh snow in a pine forest at dawn, its breath steaming in the cold air, the camera tracking alongside
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.1-vace-1.3b-fast","modality":"video","params":{"prompt":"A red fox trots through fresh snow in a pine forest at dawn, its breath steaming in the cold air, the camera tracking alongside"},"seed":59823}'Understand
Ask about a photo, a clip or a recording —
nothing leaves the machine
Models that look or listen, and answer in text.

Tomatoes
Basil
Spaghetti
Parmesan cheese
Garlic clove
List every ingredient you can see on this counter, one per line.
Qwen3-VL 8B · M5 Pro 24 GB · 4 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"qwen3vl:8b","modality":"text","params":{"prompt":"List every ingredient you can see on this counter, one per line.","max_tokens":256},"input_artifacts":["<artifact id>"]}'
- All 24 garden plots are let; six households wait for a plot.
- Total budget is £700, broken down into seeds (£420), tank (£185), and tools (£95).
- Upcoming dates: Seed swap on Sunday 5 April (11:00–13:00); open day Saturday 20 June; next meeting Saturday 9 May at 10:00.
Read this page and summarise it in three short bullet points. Keep every date and amount exactly as written.
GLM-4.6V Flash · M5 Pro 24 GB · 15 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"glm-4.6v:flash","modality":"text","params":{"prompt":"Read this page and summarise it in three short bullet points. Keep every date and amount exactly as written.","max_tokens":400},"input_artifacts":["<artifact id>"]}'A small, weathered boat drifts slowly through shallow water on a wet pavement. The camera follows it as it moves forward, reflecting the warm glow of nearby lights on its surface and the surrounding area.
What happens in this video? Two sentences.
Qwen3-VL 8B · M5 Pro 24 GB · 11 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"qwen3vl:8b","modality":"text","params":{"prompt":"What happens in this video? Two sentences.","max_tokens":256},"input_artifacts":["<artifact id>"]}'Welcome back. Today, we're looking at three small ways to make your morning coffee better. Starting with the water: fresh, filtered, and just off the boil.
Transcribe this audio.
Qwen3-ASR 1.7B · M5 Pro 24 GB · 2 s
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"qwen3-asr:1.7b","modality":"text","params":{"prompt":"Transcribe this audio.","max_tokens":256},"input_artifacts":["<artifact id>"]}'Dashboard
One dashboard
for every machine
Real screens of the dashboard, each on the machines its caption names.
Across machines
One model too big for one Mac —
and two jobs at once
Two Macs, sharing one model, then each making something of its own.
Qwen3.6 35B-A3B, a 35-billion-parameter mixture-of-experts model neither Mac can run alone, answering at 9.3 tokens/s — faster than you read.
- M5 Pro, 24 GB of memory, holds 11.5 GB of the model: layers 0 to 19 of 40, and runs the orchestrator.
- M4 Pro, 24 GB of memory, holds 9.9 GB of the model: layers 20 to 39 of 40.
- M5 Pro and M4 Pro are linked over LAN Wi-Fi.
- Together they run the model at 9.3 tokens/s.
21 GB model24 GB each9.3 tokens/s
A 24 GB Mac can give one model about 19 GB, so each holds half of its layers.
Neural networks are fundamentally structured as sequential computation graphs where each layer’s output directly becomes the next layer’s input, creating clear computational boundaries that naturally support partitioning across devices. By assigning non-overlapping blocks of layers to separate computers, each machine can independently execute its forward and backward passes while exchanging only intermediate activations and gradients through network links. Modern distributed deep learning frameworks automate this pipeline parallelism by optimizing data routing, overlapping communication with computation, and load-balancing the workload to minimize latency. This layer-wise splitting makes it possible to train or run models that exceed a single machine’s memory or compute capacity by efficiently pooling hardware resources across multiple systems.
Explain in four sentences why a model can be split by layers across two computers.
Qwen3.6 35B-A3B · M5 Pro + M4 Pro, 24 GB each, Wi-Fi
How it's made
The orchestrator was started with MAX_TOKENS=8192.
curl -X POST http://localhost:8001/v1/chat/completions \
-H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"Explain in four sentences why a model can be split by layers across two computers."}],"max_tokens":2048,"model":"qwen3.6:35b-a3b"}'Two machines, two jobs, at once
Two jobs, one per Mac, ran side by side — neither waited for the other: M5 Pro 24 GB made a picture in 69 s while M4 Pro 24 GB rendered a 2 s clip in 11.5 min.
M4 Pro 24 GB: Wan 2.2 TI2V 5B Turbo, 11.5 min
M5 Pro 24 GB: Z-Image Turbo, 69 s
Gentle waves rolling onto a sandy beach at sunset, soft foam, a wide shot.
2 s clip · 1280×704 · Wan 2.2 TI2V 5B Turbo · M4 Pro 24 GB · 11.5 min
Prompt
gentle waves rolling onto a sandy beach at sunset, soft foam, warm light, wide shot
How it's made
The orchestrator was started with DPPAN_VISUAL_NODES=on.
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"wan2.2-ti2v-5b-turbo","modality":"video","params":{"prompt":"gentle waves rolling onto a sandy beach at sunset, soft foam, warm light, wide shot"},"seed":152}'
A lighthouse keeper's desk by a round window: telescope, logbook, tea, a stormy sea.
Z-Image Turbo · M5 Pro 24 GB · 69 s
Prompt
a lighthouse keeper's desk by a round window, brass telescope, logbook and a mug of tea, stormy sea outside, photograph
How it's made
curl -X POST http://localhost:8001/v1/generations \
-H 'Content-Type: application/json' \
-d '{"model":"z-image-turbo","modality":"image","params":{"prompt":"a lighthouse keeper'\''s desk by a round window, brass telescope, logbook and a mug of tea, stormy sea outside, photograph"},"seed":151}'




