Z.ai · Chat and reasoning · 31.2B parameters · mit
GLM-4.7 Flash
Needs 42 GB at FP8 / INT8. The smallest real configuration today is 2× Tesla V100 32GB at Vast.ai, $0.14 per hour, about $99 a month at 730 hours.
Parameters31.2Bdense
Context198k tokensmaximum published
Checkpoint62 GBBF16, as published
Licensemitcheck before commercial use
Released—documented date
Smallest configuration · FP8 / INT8
2× Tesla V100 32GB64 GB VRAM · 64% used
- Per hour
- $0.14
- Per month
- $99
- Cheapest at
Vast.ai
Memory · 3 precisions · 5 context lengthsHow much VRAM it needs
Weights plus runtime overhead, then the KV cache that grows with the context you serve.
| Precision | Weights | Weights + headroom | 4k context | 32k context | 128k context | 198k context |
|---|---|---|---|---|---|---|
| BF16published checkpoint | 62 GB | 75 GB | 76 GB | 77 GB | 83 GB | 86 GB |
| FP8 / INT8published checkpoint | 34 GB | 42 GB | 42 GB | 43 GB | 49 GB | 53 GB |
| INT4 / MXFP4published checkpoint | 20 GB | 24 GB | 25 GB | 26 GB | 32 GB | 35 GB |
| KV cacheone sequence, FP16 cache, included above | — | — | 1 GB | 2 GB | 8 GB | 11 GB |
Cross-checked on Hugging Face Oct 28, 2026: 31.2B parameters in the safetensors headers, 62.4 GB of weight files, attention layout from config.json (47 layers, latent rank 512). Method: weights as published at the native precision, verified on Hugging Face 2026-10-28; other precisions estimated at 1 (8-bit) and 0.6 (4-bit) bytes per parameter; × 1.2 runtime headroom; KV cache = 1152 bytes per token per layer × 47 full-attention layers (FP16 cache, one sequence).config.json ↗
Smallest configuration per GPU · 109 GPUs · cheapest eligible on-demand offer per GPU count, every provider on fileWhat it costs to run, GPU by GPU
The fewest GPUs of one type whose total memory fits the model at short context, priced from the cheapest eligible on-demand offer for exactly that count (45-day window, newest reading per provider), with the next providers beside it. Monthly = 730 hours.
| GPU | Smallest config | VRAM used | $ / hour | $ / month | Cheapest at |
|---|---|---|---|---|---|
| Tesla V100 32GBBest fit | 2× 32 GB64 GB total | 64% | $0.14 | $99 | Vast.aiTrust 78 · read Nov 2 · receipt · also Cyfuture AI $1.13, io.net $2.52 |
| RTX 3060 | 4× 12 GB48 GB total | 86% | $0.16 | $118 |  Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.21, Edgevana $0.31 |
| Tesla P100 | 4× 16 GB64 GB total | 64% | $0.19 | $142 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.43 |
| RTX 3090 | 2× 24 GB48 GB total | 86% | $0.21 | $157 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.24, Lium $0.44 |
| Quadro RTX 8000 | 1× 48 GB48 GB total | 86% | $0.23 | $167 | Vast.aiTrust 78 · read Nov 2 · receipt · also TensorDock $0.32 |
| Titan RTX | 2× 24 GB48 GB total | 86% | $0.27 | $196 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.52 |
| Titan Xp | 4× 12 GB48 GB total | 86% | $0.27 | $196 | Vast.aiTrust 78 · read Nov 2 · receipt |
| L40 | 1× 48 GB48 GB total | 86% | $0.27 | $196 | Vast.aiTrust 78 · read Nov 2 · receipt · also io.net $0.66, Runpod $0.69 |
| RTX 2080 Ti | 4× 11 GB44 GB total | 94% tight | $0.28 | $202 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX A6000 | 1× 48 GB48 GB total | 86% | $0.33 | $241 |  GPURentonot scored · read Nov 2 · receipt · also gcube $0.34, Thunder Compute $0.35 |
| A40 | 1× 48 GB48 GB total | 86% | $0.35 | $255 | GPURentonot scored · read Nov 2 · receipt · also Vast.ai $0.48, Exoscale $1.05 |
| RTX 6000 | 2× 24 GB48 GB total | 86% | $0.35 | $256 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.40 |
| RTX 5060 Ti | 4× 16 GB64 GB total | 64% | $0.35 | $258 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.40, Edgevana $0.68 |
| RTX A4000 | 4× 16 GB64 GB total | 64% | $0.36 | $266 | Vast.aiTrust 78 · read Nov 2 · receipt · also AtmosCompute $0.66, Spheron $3.84 |
| RTX A5000 | 2× 24 GB48 GB total | 86% | $0.38 | $274 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.74, Airon AI $0.90 |
| RTX 4070 Super | 4× 12 GB48 GB total | 86% | $0.38 | $274 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.58, AtmosCompute $0.69 |
| Tesla V100 16GB | 4× 16 GB64 GB total | 64% | $0.38 | $275 | Vast.aiTrust 78 · read Nov 2 · receipt · also CloudRift $1.00, AtmosCompute $1.30 |
| RTX 6000 Ada | 1× 48 GB48 GB total | 86% | $0.40 | $290 | |
| RTX PRO 5000 | 1× 48 GB48 GB total | 86% | $0.40 | $294 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.57, Clore.ai $0.63 |
| Quadro RTX 5000 | 4× 16 GB64 GB total | 64% | $0.40 | $294 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 3090 Ti | 2× 24 GB48 GB total | 86% | $0.42 | $304 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.48, Edgevana $0.59 |
| L40S | 1× 48 GB48 GB total | 86% | $0.42 | $304 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.50, GPUBank $0.55 |
| RTX PRO 4000 | 2× 24 GB48 GB total | 86% | $0.45 | $332 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.62, Clore.ai $0.75 |
| A800 80GB PCIe | 1× 80 GB80 GB total | 52% | $0.47 | $342 | Vast.aiTrust 78 · read Nov 2 · receipt |
| The 24 cheapest of 72 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
Throughput estimates · Oct 4, 2026Tokens per second, and what a million costs
Chat and reasoning. Speed estimated from memory bandwidth and model size, never measured. The cost per 1M output tokens is an estimate at the prices read on Oct 4, 2026; it is not recomputed here.
| GPU | Cards | Precision | tok/s per user | per 1M output tokens · estimate |
|---|---|---|---|---|
| RTX 4090 | 2 | FP8 | 49.5 | $0.108 |
| MI300X | 1 | FP8 | 123 | $0.143 |
| RTX 5090 | 2 | FP8 | 63.1 | $0.154 |
| L40S | 1 | FP8 | 61.3 | $0.166 |
| H100 SXM | 1 | FP8 | 132 | $0.177 |
| RTX PRO 6000 | 1 | FP8 | 99.3 | $0.186 |
| H200 | 1 | FP8 | 148 | $0.190 |
| H100 PCIe | 1 | FP8 | 101 | $0.194 |
| B200 | 1 | FP8 | 200 | $0.200 |
| A100 80GB | 1 | BF16 | 68.5 | $0.273 |
cards, precision and speed estimated from memory bandwidth and model size; cost recomputed by the site from verified prices · cheapest per output at the prices of Oct 4, 2026: RTX 4090Why estimates, and their limits →
GLM-4.7 Flash · related