Models / Llama 3.3 70B Instruct / Calculator

Llama 3.3 70B Instruct cost & VRAM calculator

What does Llama 3.3 70B Instruct cost to run for your workload, and can you run it on your own hardware? Set your workload below - we compute per-provider API cost live and tell you honestly whether local hardware can run it.

Your workload

Casual Power user Team of 5 Codebase task

Llama 3.3 70B Instruct runs an always-on thinking mode. Reasoning (thinking) tokens are billed at the output rate ($15.00/M), so count them here to see the thinking portion of your bill.

API cost for your workload

Provider Rate ($/1M) Monthly cost
DeepInfra $0.10 in · $0.32 out $0.0 cheapest
Novita $0.14 in · $0.40 out $0.0 cheapest
AkashML $0.20 in · $0.52 out · $0.10 cache $0.0 cheapest
Parasail $0.22 in · $0.50 out · $0.11 cache $0.01
Cloudflare $0.29 in · $2.25 out $0.01
SambaNova $0.45 in · $0.90 out $0.01
Groq $0.59 in · $0.79 out · $0.30 cache $0.01
CoreWeave $0.71 in · $0.71 out · $0.71 cache $0.02
Google $0.72 in · $0.72 out $0.02
Together AI may be stale $0.88 in · $0.88 out $0.02
Fireworks AI may be stale $0.90 in · $0.90 out $0.02

Monthly cost is an estimate from list prices and your workload - verify against the provider before committing. Cached fraction applies the cache rate to that share of input.

Can you run it locally?

YES - IT RUNS ON HARDWARE YOU CAN BUY

Llama 3.3 70B Instruct fits on individual rigs. Here's the hardware that runs it comfortably; the full per-quant VRAM matrix and tok/s estimates live on the model page.

4x H100 80GB (320GB) 320GB Q2_K
342.7t/s
NVIDIA DGX Station 748GB 748GB Q2_K
204.6t/s
8x RTX 3090 rack (192GB) 192GB Q2_K
191.5t/s
4x RTX 5090 (128GB) 128GB Q2_K
183.3t/s
AMD Instinct MI300X (192GB) 192GB Q2_K
136.2t/s
4x RTX 4090 (96GB) 96GB Q2_K
103.1t/s
2x RTX 5090 (64GB) 64GB Q2_K
91.7t/s
2x RTX 3090 (48GB) 48GB Q2_K
47.9t/s
Single RTX 5090 (32GB) 32GB Q2_K
45.8t/s
RTX PRO 6000 Blackwell (96GB) 96GB Q2_K
45.8t/s
Mac Studio M4 Ultra 192GB 192GB Q2_K
30.5t/s
Mac Studio M4 Ultra 512GB 512GB Q2_K
30.5t/s
MacBook Pro M5 Max 128GB 128GB Q2_K
17.1t/s
Dual EPYC 9004 + 768GB DDR5-4800 768GB Q2_K
11.8t/s
DGX Spark 128GB unified 128GB Q2_K
7.0t/s
Ryzen AI Max+ 395 128GB 128GB Q2_K
6.6t/s
Jetson AGX Orin 64GB 64GB Q2_K
5.2t/s
Epyc + 512GB DDR4-3200 + 2x RTX 3090 560GB Q2_K
5.2t/s
Epyc + 512GB DDR4-2400 + 2x RTX 3090 560GB Q2_K
3.9t/s
Q2_K
21.5GB weights 21.5GB min 25.0GB rec
Q3_K_M
30.0GB weights 30.0GB min 34.0GB rec
Q4_K_M
42.0GB weights 42.0GB min 48.0GB rec
Q8_0
74.0GB weights 74.0GB min 80.0GB rec

Want the own-vs-rent payback (capex vs your monthly API bill)? The budget tool computes that against representative rigs.

Full model card API pricing table Generic token calculator