portfolio/glm52.html at main · RizzuAI/portfolio · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
<!DOCTYPE html>
<html lang="it">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>GLM-5.2 · 744B MoE</title>
<style>
*{margin:0;padding:0;box-sizing:border-box}
html{scroll-behavior:smooth}
body{
  font-family:system-ui,-apple-system,'Helvetica Neue',sans-serif;
  background:#fafafa;color:#1d1d1f;
  line-height:1.47059;
  -webkit-font-smoothing:antialiased;
  overflow-x:hidden;
}
.w{max-width:980px;margin:0 auto;padding:0 22px}

/* HERO - full screen, no background noise */
.hero{
  min-height:100vh;display:flex;flex-direction:column;justify-content:center;
  background:#fff;position:relative;overflow:hidden;
}
.hero .w{position:relative;z-index:2}
.hero .sticker{
  position:absolute;top:50%;right:-80px;transform:translateY(-50%);
  font-size:clamp(300px,40vw,700px);
  font-weight:900;color:#f0f0f2;letter-spacing:-.06em;user-select:none;
  line-height:1;
  z-index:1;
  animation:float 8s ease-in-out infinite;
}
@keyframes float{
  0%,100%{transform:translateY(-50%) translateX(0)}
  50%{transform:translateY(-55%) translateX(-20px)}
}
.hero .eyebrow{
  font-size:14px;font-weight:500;color:#86868b;
  letter-spacing:-.01em;margin-bottom:24px;
}
.hero h1{
  font-size:clamp(40px,6vw,76px);
  font-weight:700;letter-spacing:-.03em;
  line-height:1.03;margin-bottom:16px;
  max-width:780px;
  animation:slideUp 1s cubic-bezier(.16,1,.3,1) both;
}
.hero h1 span{background:linear-gradient(135deg,#14b8a6,#0891b2);-webkit-background-clip:text;background-clip:text;color:transparent}
.hero p{
  font-size:20px;color:#86868b;max-width:520px;line-height:1.5;margin-bottom:36px;
  animation:slideUp 1s .15s cubic-bezier(.16,1,.3,1) both;
}
.cta-row{display:flex;gap:12px;flex-wrap:wrap;animation:slideUp 1s .3s cubic-bezier(.16,1,.3,1) both}
.btn-pri,.btn-sec{
  display:inline-flex;align-items:center;gap:6px;
  padding:10px 22px;border-radius:980px;
  font-size:14px;font-weight:500;text-decoration:none;cursor:pointer;transition:all .2s;
}
.btn-pri{background:#14b8a6;color:#fff;border:none}
.btn-pri:hover{background:#0d9488;transform:scale(1.02)}
.btn-sec{border:1px solid #d4d4d8;color:#1d1d1f;background:transparent}
.btn-sec:hover{border-color:#a1a1aa}

@keyframes slideUp{
  from{opacity:0;transform:translateY(40px)}
  to{opacity:1;transform:none}
}

/* METRICS - horizontal scroll of numbers */
.metrics{
  display:flex;gap:0;margin-top:48px;
  animation:slideUp 1s .45s cubic-bezier(.16,1,.3,1) both;
  border-top:1px solid #e8e8ed;padding-top:32px;
}
.metric{flex:1}
.metric .v{font-size:24px;font-weight:700;letter-spacing:-.02em}
.metric .l{font-size:13px;color:#86868b;margin-top:2px}
.metric+.metric{border-left:1px solid #e8e8ed;padding-left:24px;margin-left:24px}

/* SECTION HEADERS - minimal */
.sn{padding:120px 0;position:relative}
.sn.light{background:#fff}
.sn.dark{background:#1d1d1f;color:#f5f5f7}
.s-label{
  font-size:11px;font-weight:600;text-transform:uppercase;
  letter-spacing:.12em;color:#14b8a6;margin-bottom:8px;
}
.s-title{
  font-size:clamp(32px,4vw,52px);
  font-weight:700;letter-spacing:-.025em;
  line-height:1.05;margin-bottom:10px;
}
.s-sub{
  font-size:18px;color:#86868b;max-width:520px;line-height:1.5;
}

/* ARCHITECTURE - timeline style */
.timeline{position:relative;margin-top:60px}
.timeline::before{
  content:'';position:absolute;left:20px;top:0;bottom:0;
  width:2px;background:#e8e8ed;
}
.tl-item{position:relative;padding:0 0 32px 56px}
.tl-item:last-child{padding-bottom:0}
.tl-dot{
  position:absolute;left:14px;top:4px;
  width:14px;height:14px;border-radius:50%;
  background:#14b8a6;border:3px solid #fff;
}
.tl-item .tag{
  display:inline-block;font-size:11px;font-weight:600;
  color:#14b8a6;font-family:ui-monospace,monospace;
  margin-bottom:2px;
}
.tl-item h3{font-size:17px;font-weight:600;margin-bottom:2px}
.tl-item p{font-size:14px;color:#6b7280;line-height:1.5;max-width:500px}

/* FEATURES - staggered cards */
.feat-grid{
  display:grid;grid-template-columns:1fr 1fr;gap:1px;
  background:#e8e8ed;border-radius:16px;overflow:hidden;margin-top:40px;
}
.feat{background:#fff;padding:32px}
.feat h3{font-size:15px;font-weight:600;margin-bottom:4px}
.feat p{font-size:14px;color:#6b7280;line-height:1.5}
.feat .num{
  font-size:11px;font-weight:700;color:#14b8a6;
  font-family:ui-monospace,monospace;margin-bottom:6px;
}

/* COLIBRÌ - split */
.cb-wrap{display:grid;grid-template-columns:1fr 1fr;gap:48px;margin-top:48px}
.cb-text p{font-size:15px;color:#a1a1aa;line-height:1.6}
.res-bars{margin-top:24px}
.rb{margin-bottom:14px}
.rb .l{display:flex;justify-content:space-between;font-size:12px;margin-bottom:4px}
.rb .l span:last-child{color:#14b8a6;font-weight:500}
.rb .t{height:4px;background:#3f3f46;border-radius:2px;overflow:hidden}
.rb .f{height:100%;background:linear-gradient(90deg,#14b8a6,#0891b2);border-radius:2px}

.term{
  background:#09090b;border:1px solid #27272a;
  border-radius:12px;font-family:ui-monospace,monospace;
  font-size:12px;line-height:1.8;overflow:hidden;
}
.term-h{display:flex;gap:6px;padding:10px 14px;border-bottom:1px solid #27272a;background:#18181b;color:#71717a}
.term-h .d{width:10px;height:10px;border-radius:50%;background:#3f3f46}
.term-b{padding:14px;color:#71717a}
.term-b .ok{color:#14b8a6}
.term-b .pr{color:#fafafa}
.term-b .i{color:#52525b}

/* BENCHMARKS - big numbers */
.bn-wrap{display:grid;grid-template-columns:repeat(3,1fr);gap:24px;margin-top:40px}
.bn-card{
  background:#fff;border:1px solid #e8e8ed;border-radius:16px;
  padding:32px;text-align:center;
}
.bn-card .name{font-size:12px;color:#86868b;font-family:ui-monospace,monospace;margin-bottom:6px}
.bn-card .score{font-size:48px;font-weight:700;letter-spacing:-.03em;line-height:1}
.bn-card .score .p{font-size:20px;color:#a1a1aa;font-weight:400}
.bn-card .bar{height:3px;background:#e8e8ed;border-radius:1.5px;margin-top:16px;overflow:hidden}
.bn-card .bar i{display:block;height:100%;background:#14b8a6;border-radius:1.5px}

/* TABLE - clean */
.cmp{margin-top:60px}
.cmp h3{font-size:16px;font-weight:600;margin-bottom:16px}
table{width:100%;border-collapse:collapse;font-size:13px}
th,td{padding:12px 14px;text-align:left;border-bottom:1px solid #e8e8ed}
th{font-size:11px;font-weight:500;color:#86868b;text-transform:uppercase;letter-spacing:.06em}
td{color:#6b7280}
td.n{color:#1d1d1f;font-weight:500}
td.n .s{font-size:11px;color:#86868b;font-weight:400}
.mf{font-family:ui-monospace,monospace;font-size:12px}
tr:last-child td{border-bottom:none}

/* PRICING - simple */
.pg{display:grid;grid-template-columns:repeat(3,1fr);gap:1px;background:#e8e8ed;border-radius:16px;overflow:hidden;margin-top:40px}
.pc{background:#fff;padding:32px}
.pc .t{font-size:13px;color:#86868b;margin-bottom:4px}
.pc .a{font-size:32px;font-weight:700;letter-spacing:-.02em;margin-bottom:2px}
.pc .a .u{font-size:13px;color:#86868b;font-weight:400}
.pc .p{font-size:12px;color:#a1a1aa;margin-bottom:14px}
.pc ul{list-style:none}
.pc li{font-size:13px;color:#6b7280;padding:2px 0}

/* CTA */
.cta-sn{padding:120px 0;text-align:center;background:linear-gradient(180deg,#fff,#f5f5f7)}
.cta-sn h2{margin-bottom:8px}
.cta-sn p{margin:0 auto 28px}
.cta-sn .cta-row{justify-content:center}

/* FOOTER */
footer{padding:40px 0;font-size:12px;color:#86868b;border-top:1px solid #e8e8ed}
footer .w{display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:16px}
footer a{color:#86868b;text-decoration:none;margin-left:20px;transition:color .2s}
footer a:hover{color:#1d1d1f}

/* REVEAL - working properly */
.r{opacity:0;transform:translateY(30px);transition:opacity .8s cubic-bezier(.16,1,.3,1),transform .8s cubic-bezier(.16,1,.3,1)}
.r.go{opacity:1;transform:none}
.r1{transition-delay:.12s}.r2{transition-delay:.24s}.r3{transition-delay:.36s}.r4{transition-delay:.48s}

/* MOBILE */
@media(max-width:768px){
  .hero h1{font-size:36px}
  .metrics{flex-direction:column;gap:16px}
  .metric+.metric{border-left:none;padding-left:0;margin-left:0;padding-top:16px;border-top:1px solid #e8e8ed}
  .sn{padding:80px 0}
  .feat-grid,.cb-wrap,.pg{grid-template-columns:1fr}
  .bn-wrap{grid-template-columns:1fr}
  .hero .sticker{display:none}
}
</style>
</head>
<body>

<header class="hero">
  <div class="sticker">744</div>
  <div class="w">
    <div class="eyebrow">Zhipu AI · 2026</div>
    <h1>744 miliardi.<br><span>1 milione</span> di contesto.</h1>
    <p>Mixture-of-Experts open source. 40B attivi per token. Competitivo con i migliori modelli chiusi. Gira su CPU con 25 GB di RAM.</p>
    <div class="cta-row">
      <a href="#bench" class="btn-pri">Vedi benchmark →</a>
      <a href="#colibri" class="btn-sec">colibrì runtime</a>
    </div>
    <div class="metrics">
      <div class="metric"><div class="v">744B</div><div class="l">Parametri totali</div></div>
      <div class="metric"><div class="v">40B</div><div class="l">Attivi per token</div></div>
      <div class="metric"><div class="v">1M</div><div class="l">Token di contesto</div></div>
      <div class="metric"><div class="v">256</div><div class="l">Esperti MoE</div></div>
    </div>
  </div>
</header>

<section class="sn light" id="arch">
  <div class="w">
    <div class="r"><div class="s-label">Architettura</div></div>
    <h2 class="s-title r r1">MLA + MoE,<br>75 layer di profondità.</h2>
    <p class="s-sub r r2">Attention compressa, routing sigmoid, decoding speculativo. I pesi int4 in 9.9 GB di RAM.</p>
    <div class="timeline r r2">
      <div class="tl-item">
        <div class="tl-dot"></div>
        <div class="tag">MLA</div>
        <h3>Multi-Head Latent Attention</h3>
        <p>KV-cache compressa: 576 float per token invece di 32.768. Contesto da 1M con memoria ridotta del 57×.</p>
      </div>
      <div class="tl-item">
        <div class="tl-dot" style="background:#0891b2"></div>
        <div class="tag">MoE · 256 experts</div>
        <h3>Sigmoid Router</h3>
        <p>Ogni esperto pesato indipendentemente. Attivazione stabile, bilanciamento naturale. Stile DeepSeek-V3.</p>
      </div>
      <div class="tl-item">
        <div class="tl-dot" style="background:#0e7490"></div>
        <div class="tag">MTP</div>
        <h3>Multi-Token Prediction</h3>
        <p>Decoding speculativo nativo. Accept rate 39–59%. Fino a 2.8 token per forward.</p>
      </div>
      <div class="tl-item">
        <div class="tl-dot" style="background:#155e75"></div>
        <div class="tag">int4</div>
        <h3>Quantizzazione nativa</h3>
        <p>9.9 GB residenti. Inferenza su CPU consumer. Runtime colibrì: un file C, zero dipendenze.</p>
      </div>
    </div>
  </div>
</section>

<section class="sn" id="feat" style="background:#f5f5f7">
  <div class="w">
    <div class="r"><div class="s-label">Capacità</div></div>
    <h2 class="s-title r r1">Sei ambiti.</h2>
    <div class="feat-grid r r2">
      <div class="feat"><div class="num">01</div><h3>Coding</h3><p>HumanEval 84.7%. Refactoring, debug, completamento repository con contesto 1M.</p></div>
      <div class="feat"><div class="num">02</div><h3>Tool calling</h3><p>Funzioni strutturate, agenti multi-step, chiamate parallele affidabili.</p></div>
      <div class="feat"><div class="num">03</div><h3>Bilingue</h3><p>Cinese e inglese nativi. Comprensione culturale, non traduzione automatica.</p></div>
      <div class="feat"><div class="num">04</div><h3>Matematica</h3><p>GSM8K 91.3%. Deduzioni multi-passo, verifica, problemi lunghi.</p></div>
      <div class="feat"><div class="num">05</div><h3>Open source</h3><p>Apache 2.0 (codice), MIT (pesi). Self-hosting, modifica, audit libero.</p></div>
      <div class="feat"><div class="num">06</div><h3>Inferenza leggera</h3><p>25 GB RAM su CPU consumer via colibrì. Nessuna GPU obbligatoria.</p></div>
    </div>
  </div>
</section>

<section class="sn dark" id="colibri">
  <div class="w">
    <div class="r"><div class="s-label" style="color:#14b8a6">colibrì</div></div>
    <h2 class="s-title r r1" style="color:#fafafa">CPU, niente GPU.</h2>
    <div class="cb-wrap">
      <div class="cb-text r r2">
        <p>colibrì carica GLM-5.2 in 9.9 GB di RAM e lo esegue su CPU x86. Un file C, zero dipendenze. 25 GB di RAM per inferenza completa.</p>
        <div class="res-bars">
          <div class="rb"><div class="l"><span>RAM necessaria</span><span>25 GB</span></div><div class="t"><div class="f" style="width:78%"></div></div></div>
          <div class="rb"><div class="l"><span>Footprint RAM</span><span>9.9 GB</span></div><div class="t"><div class="f" style="width:31%"></div></div></div>
        </div>
      </div>
      <div class="term r r2">
        <div class="term-h"><span class="d"></span><span class="d"></span><span class="d"></span><span style="margin-left:8px">colibrì · glm-5.2</span></div>
        <div class="term-b">
          <div><span class="i">$</span> <span class="pr">colibri load glm-5.2-int4</span></div>
          <div class="ok">✓ 9.9 GB residenti · 256 esperti</div>
          <div class="ok">✓ contesto: 1.048.576 token</div>
          <div class="ok">✓ pronto in 3.2 s</div>
          <br>
          <div><span class="i">$</span> <span class="pr">colibri chat</span></div>
          <div class="ok">✓ 247 tok/s · CPU 8-core</div>
        </div>
      </div>
    </div>
  </div>
</section>

<section class="sn light" id="bench">
  <div class="w">
    <div class="r"><div class="s-label">Benchmark</div></div>
    <h2 class="s-title r r1">All'altezza dei migliori.</h2>
    <p class="s-sub r r2">Precisione int4 su suite standard.</p>
    <div class="bn-wrap r r2">
      <div class="bn-card"><div class="name">MMLU</div><div class="score">89.2<span class="p">%</span></div><div class="bar"><i style="width:89.2%"></i></div></div>
      <div class="bn-card"><div class="name">HumanEval</div><div class="score">84.7<span class="p">%</span></div><div class="bar"><i style="width:84.7%"></i></div></div>
      <div class="bn-card"><div class="name">GSM8K</div><div class="score">91.3<span class="p">%</span></div><div class="bar"><i style="width:91.3%"></i></div></div>
    </div>
    <div class="cmp r r2">
      <h3>GLM-5.2 · DeepSeek V4 Flash · Grok 4.5</h3>
      <table>
        <thead><tr><th>Modello</th><th>Params</th><th>Attivi</th><th>Contesto</th><th>Licenza</th><th>MMLU</th><th>HumanEval</th><th>GSM8K</th></tr></thead>
        <tbody>
          <tr><td class="n">GLM-5.2<div class="s">Zhipu AI</div></td><td class="mf">744B</td><td class="mf">40B</td><td class="mf">1M</td><td>Apache 2.0 · MIT</td><td class="mf">89.2</td><td class="mf">84.7</td><td class="mf">91.3</td></tr>
          <tr><td class="n">DeepSeek V4 Flash<div class="s">DeepSeek</div></td><td class="mf">685B</td><td class="mf">22B</td><td class="mf">128K</td><td>MIT</td><td class="mf">87.4</td><td class="mf">82.1</td><td class="mf">89.6</td></tr>
          <tr><td class="n">Grok 4.5<div class="s">xAI</div></td><td class="mf">~620B</td><td class="mf">38B</td><td class="mf">256K</td><td>Proprietario</td><td class="mf">90.1</td><td class="mf">86.2</td><td class="mf">92.0</td></tr>
        </tbody>
      </table>
    </div>
  </div>
</section>

<section class="sn" style="background:#f5f5f7" id="price">
  <div class="w">
    <div class="r"><div class="s-label">Prezzi</div></div>
    <h2 class="s-title r r1">Tre modi.</h2>
    <div class="pg r r2">
      <div class="pc"><div class="t">API · Base</div><div class="a">€0.40<span class="u"> /M tok</span></div><div class="p">input · €1.20 output</div><ul><li>Endpoint EU</li><li>Contesto 1M</li><li>Tool calling</li><li>100 RPM</li></ul></div>
      <div class="pc"><div class="t">API · Pro</div><div class="a">€0.28<span class="u"> /M tok</span></div><div class="p">input · €0.85 output</div><ul><li>Priorità latenza</li><li>10.000 RPM</li><li>SLA 99.9%</li><li>Batch</li></ul></div>
      <div class="pc"><div class="t">Self-hosted</div><div class="a">Gratis</div><div class="p">Pesi MIT · Apache 2.0</div><ul><li>Runtime colibrì</li><li>25 GB RAM</li><li>int4</li><li>Community</li></ul></div>
    </div>
  </div>
</section>

<section class="cta-sn" id="cta">
  <div class="w">
    <h2 class="s-title" style="font-size:clamp(28px,3.5vw,42px);margin-bottom:8px">744B. 40B attivi. 1M contesto.<br>Open source, MIT.</h2>
    <div class="cta-row" style="justify-content:center">
      <a href="#" class="btn-pri">Crea chiave API</a>
      <a href="#" class="btn-sec">Scarica i pesi</a>
    </div>
  </div>
</section>

<footer>
  <div class="w">
    <div>© 2026 Zhipu AI · Apache 2.0 · MIT</div>
    <div><a href="#">Documentazione</a><a href="#">GitHub</a><a href="#">Hugging Face</a></div>
  </div>
</footer>

<script>
(()=>{
  // Scroll reveal — works on all modern browsers
  const io = new IntersectionObserver((entries) => {
    entries.forEach(e => {
      if (e.isIntersecting) {
        e.target.classList.add('go');
        io.unobserve(e.target);
      }
    });
  }, { threshold: 0.08, rootMargin: '0px 0px -20px 0px' });

  document.querySelectorAll('.r').forEach(el => io.observe(el));

  // Trigger hero animations immediately (they use CSS animation, not IntersectionObserver)
  console.log('GLM-5.2 page ready');
})();
</script>
</body>
</html>