Models / GLM 5.2 / Calculator

GLM 5.2 cost & VRAM calculator

What does GLM 5.2 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

GLM 5.2 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
Relace $0.03 in · $7.15 out · $0.03 cache $0.0 cheapest
Wafer $0.19 in · $8.00 out · $0.18 cache $0.0 cheapest
Cloudflare $0.19 in · $8.00 out · $0.19 cache $0.0 cheapest
DeepInfra $0.56 in · $1.80 out · $0.10 cache $0.01
Ambient $0.60 in · $2.00 out · $0.15 cache $0.01
StreamLake $0.61 in · $1.93 out · $0.11 cache $0.01
CoreWeave $0.62 in · $2.42 out · $0.14 cache $0.01
Novita $0.65 in · $2.04 out · $0.12 cache $0.02
DigitalOcean $0.70 in · $2.20 out · $0.10 cache $0.02
SiliconFlow $0.70 in · $2.20 out · $0.13 cache $0.02
DigitalOcean may be stale $0.70 in · $2.20 out · $0.10 cache $0.02
AtlasCloud $0.94 in · $2.95 out · $0.17 cache $0.02
Nous Portal may be stale $0.95 in · $2.99 out $0.02
Inceptron $0.99 in · $2.88 out · $0.16 cache $0.02
Decart $1.20 in · $8.00 out · $0.48 cache $0.03
Together $1.40 in · $4.40 out · $0.26 cache $0.03
Friendli $1.40 in · $4.40 out · $0.26 cache $0.03
Parasail $1.40 in · $4.40 out · $0.26 cache $0.03
Venice $1.40 in · $4.40 out · $0.26 cache $0.03
Z.ai may be stale $1.40 in · $4.40 out $0.03
Z.AI $1.40 in · $4.40 out · $0.26 cache $0.03
GMICloud $1.40 in · $4.40 out · $0.26 cache $0.03
Nebius $1.40 in · $4.40 out · $0.14 cache $0.03
Mistral $1.54 in · $4.84 out · $0.15 cache $0.04
BaseTen $2.10 in · $6.60 out · $0.21 cache $0.05
Baidu $2.25 in · $7.88 out · $0.56 cache $0.05
Alibaba $2.31 in · $7.26 out · $0.46 cache $0.06
Morph $0.57 in · $4.40 out · $0.14 cache $0.01

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

GLM 5.2 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
570.3t/s
NVIDIA DGX Station 748GB 748GB Q2_K
340.5t/s
Mac Studio M4 Ultra 512GB 512GB Q2_K
50.7t/s
Dual EPYC 9004 + 768GB DDR5-4800 768GB Q2_K
19.6t/s
Epyc + 512GB DDR4-3200 + 2x RTX 3090 560GB Q2_K
8.7t/s
Epyc + 512GB DDR4-2400 + 2x RTX 3090 560GB Q2_K
6.5t/s
Q2_K
240.0GB weights 240.0GB min 260.0GB rec
Q3_K_M
365.0GB weights 365.0GB min 400.0GB rec
Q4_K_M
410.0GB weights 410.0GB min 450.0GB rec
Q5_K_M
510.0GB weights 510.0GB min 560.0GB rec
BF16
1488.0GB weights 1488.0GB min 1600.0GB 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