Llama 3.3 70B Instruct

LicenceFACT

llama3.3

Fresh·Retrieved 25 Aug 2026

ParametersFACT

70.6B

Fresh·Retrieved 25 Aug 2026

Context lengthFACT

not returned by the registry

Evidence not verified

Downloads (30 days)FACT

301,057

Fresh·Retrieved 25 Aug 2026

LikesFACT

2,972

Fresh·Retrieved 25 Aug 2026

GatedFACT

yes

Fresh·Retrieved 25 Aug 2026

Peak VRAM for serving this model, computed by lib/deployment/hardware.ts from weights + KV cache + 15% runtime overhead. Inputs: 70.6B parameters, 8,192 tokens of context (assumed — none fetched yet), 4 concurrent requests (assumed).

• Cloud GPU instance — 1–2 × NVIDIA A100 80 GB

  GPU

  NVIDIA A100 80 GB (AWS P4d, Azure NDasr A100 v4, GCP A2)

  VRAM

  80 GB

  System RAM

  256 GB

  Storage

  2000 GB

  CPU

  24–48 vCPU

  Form factor

  Cloud instance

  Indicative costUS$2 – US$12

  indicative on-demand rental, USD per hour, Aug 2026 — verify against the provider price list for your region

  For 32B models with long context, or 70B-class models on two cards. The band covers one to two GPUs and the gap between committed-use and on-demand rates. Check that the GPU family you need exists in the region you are required to stay in before committing to a design.

  Sizing basis

  • Weights: 70.6B parameters × 2 bytes per parameter (fp16) = 141.11 GB.
  • No model configuration was supplied, so the KV cache uses the size heuristic for this band: 0.328 MB per token, from Qwen2.5-72B-Instruct (80 × 8 × 128).
  • KV cache: 0.328 MB per token × 8,192 tokens of context × 4 concurrent requests = 10.74 GB, sized for every user holding a full context at once.
  • Overhead: 15% of weights plus KV cache for the runtime, activations and memory fragmentation = 22.78 GB.
  • GB means 10⁹ bytes, matching how GPU memory is advertised.
  • KV cache held at fp16 (2 bytes per element); quantising weights does not by itself quantise the cache.
  • Peak figure, not average: prefix caching and paged attention usually keep real usage lower, while speculative decoding, CUDA graphs and a second resident model push it higher.
  • "Concurrent" counts requests generated at the same instant, not people: 4 simultaneous requests is the worst case this estimate is built on — measure yours before buying.

• On-premise dual 80 GB GPU server

  GPU

  2 × NVIDIA A100 80 GB or 2 × NVIDIA H100 80 GB with NVLink

  VRAM

  160 GB

  System RAM

  512 GB

  Storage

  8000 GB

  CPU

  Dual-socket x86 server CPU, 48+ cores total

  Form factor

  Rack server

  Indicative costUS$35,000 – US$110,000

  indicative build cost for the complete machine, USD, Aug 2026 — verify with a local supplier

  For a 70B-class model at fp16/int8 with tensor parallelism, or several mid-size models served side by side. Needs real data-centre conditions: 4–8 kW of power, front-to-back cooling and a rack. The price band is wide because A100 and H100 are several times apart.

  Sizing basis

  • Weights: 70.6B parameters × 2 bytes per parameter (fp16) = 141.11 GB.
  • No model configuration was supplied, so the KV cache uses the size heuristic for this band: 0.328 MB per token, from Qwen2.5-72B-Instruct (80 × 8 × 128).
  • KV cache: 0.328 MB per token × 8,192 tokens of context × 4 concurrent requests = 10.74 GB, sized for every user holding a full context at once.
  • Overhead: 15% of weights plus KV cache for the runtime, activations and memory fragmentation = 22.78 GB.
  • GB means 10⁹ bytes, matching how GPU memory is advertised.
  • KV cache held at fp16 (2 bytes per element); quantising weights does not by itself quantise the cache.
  • Peak figure, not average: prefix caching and paged attention usually keep real usage lower, while speculative decoding, CUDA graphs and a second resident model push it higher.
  • "Concurrent" counts requests generated at the same instant, not people: 4 simultaneous requests is the worst case this estimate is built on — measure yours before buying.

No hosted provider has been fetched for this model. OpenRouter’s catalogue is read by pnpm sync; until it runs, this is empty rather than assumed.

No published deployment stack names this model yet.

1. 01–07
   Tier 4·official repository / model cardHugging Face Hub

   Hugging Face model card data — meta-llama/Llama-3.3-70B-Instruct

   https://huggingface.co/api/models/meta-llama/Llama-3.3-70B-Instruct

   Fresh·Retrieved 25 Aug 2026sha256:8b9782788195

   7 records01 license ·02 gated ·03 downloads ·04 likes ·05 lastModified ·06 library name ·07 total