Edge cache âĸ â OPTIMIZED ⥠1.5s â 5ms cached âĸ Stealth AI Gateway
kkr-astra-cache.iamkkronly58.workers.dev astra-telemetry-service.hf.space v2-cached
| āĻāĻā§ | āĻāĻāύ | āĻāĻŋāĻāĻžāĻŦā§ |
|---|---|---|
| 5-8 āϏā§āĻā§āύā§āĻĄ | 1.4-1.6s (āĻĒā§āϰāĻĨāĻŽāĻŦāĻžāϰ) | 1B āĻŽāĻĄā§āϞ + max_tokens 150 |
| āĻĒā§āϰāϤāĻŋāĻŦāĻžāϰ 5s | 5-24ms (cached) | Cloudflare Edge Cache 60s |
| 512 tokens | 150 tokens default | āĻĻā§āϰā§āϤ āĻā§āύāĻžāϰā§āĻļāύ |
Same query 60 āϏā§āĻā§āύā§āĻĄā§āϰ āĻŽāϧā§āϝ⧠āĻāĻŦāĻžāϰ āĻāϰāϞ⧠X-Cache: HIT āĻāĻŦāĻ 5ms āĻ response! quotes āϏāĻŦ double quotes āĻ āϰāĻžāĻāĻž āĻšāϝāĻŧā§āĻā§āĨ¤
{
"status": "ok",
"query": "hello ai",
"response": "How can I assist you today?",
"model": "meta-llama/Llama-3.2-1B-Instruct",
"latency_ms": 5,
"cached": true,
"developer": "Kaustav Kanti Ray @iamkkronly"
}
| Short Code | Model ID | Size | Speed | āϧāϰāύ |
|---|---|---|---|---|
llama-1b | meta-llama/Llama-3.2-1B-Instruct | 1B | 1.5s ⥠| Fastest default |
llama-3b | meta-llama/Llama-3.2-3B-Instruct | 3B | 3.9s | Permissive |
stheno-8b | Sao10K/L3-8B-Stheno-v3.2 | 8B | 1.7s đĨ | Uncensored |
hermes-8b | NousResearch/Hermes-3-Llama-3.1-8B | 8B | 5.7s | Uncensored Hermes |
dolphin-24b | CognitiveComputations/dolphin-mistral-24b-venice-edition | 24B | 1.7s | Uncensored Dolphin |
gemma-4b | google/gemma-3-4b-it | 4B | 5.3s | Balanced |
qwen-4b | Qwen/Qwen3-4B-Instruct-2507 | 4B | 5.2s | Qwen best |
āĻŦā§āϝāĻŦāĻšāĻžāϰ: /api?chat=hello&model=stheno-8b āĻŦāĻž &model=dolphin-24b â āϏāĻŦ double quotes āĻ JSON āϰāĻŋāĻāĻžāϰā§āύ āĻšāĻŦā§āĨ¤
| āĻāĻžāϝāĻŧāĻāĻž | āĻŦā§āϝāĻŦāĻšāĻžāϰ | āϞāĻŋāĻŽāĻŋāĻ | āĻŦāĻžāĻāĻŋ |
|---|---|---|---|
| HF Space (static) | 6.98 KB (5 files) | 300 MB | 99.9% āĻĢāĻžāĻāĻāĻž |
| ââ index.html | 4.9 KB | - | - |
| ââ README.md | 0.17 KB | - | - |
| ââ api.html | 0.14 KB | - | - |
| Cloudflare Worker | 10.1 KB script | 1 MB | 99% āĻĢāĻžāĻāĻāĻž |
| ââ Memory | ~5-10 MB used | 128 MB | 92% āĻĢāĻžāĻāĻāĻž |
| ââ CPU | 5-20ms per req | 50ms | 60% āĻĢāĻžāĻāĻāĻž |
| ââ Cache | 60s edge cache | - | HIT = 5ms |
| ââ Hardware | Static (no GPU, no RAM) - āϏāĻŽā§āĻĒā§āϰā§āĻŖ āĻĢā§āϰāĻŋ | ||
đĄ HF Space static āĻšāĻāϝāĻŧāĻžāϝāĻŧ āĻā§āύ⧠GPU/RAM āϞāĻžāĻā§ āύāĻž, Cloudflare Worker edge āĻ āĻāϞ⧠â āϤāĻžāĻ āĻāϤ āĻāĻŽ memory āĻŦā§āϝāĻŦāĻšāĻžāϰāĨ¤