GLM-4.7-Flash · 10 GPUs measured first-party · llama.cpp · Updated October 2026
What GPU Do You Need for GLM-4.7-Flash?
GLM-4.7-Flash on 10 GPUs, measured first-party: NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads at 215.6 tok/s, L4 trails at 75 tok/s, and it peaked at 19GB of VRAM.
Benchmarked weights: unsloth/GLM-4.7-Flash-GGUF
NVIDIA RTX PRO 6000 Blackwell Workstation Edition
215.6 tok/s on GLM-4.7-Flash, the ceiling. Measured on our bench. 96GB of VRAM, $8,565 at launch.
- 215.6 tok/s on GLM-4.7-Flash
- 96GB, clears the GLM-4.7-Flash floor
- Rentable by the hour rather than bought
- 600W board rating
- Datacenter or workstation hardware, not a retail purchase
What GPU Do You Need for GLM-4.7-Flash?, tok/s by GPU
Efficiency: tok/s per 100W drawn
Power is the average pulled during the run, sampled at 1Hz. The fastest card is often not the one here, and for anything left running this is the number that shows up on the bill.
Value: tok/s per $1,000 of MSRP
Launch price, not street price, so it ages. A speed leaderboard always crowns the most expensive card; this is the counterweight.
GLM-4.7-Flash. Measured tokens per second by GPU
| GPU | tok/s | Prompt t/s | tok/W | Avg power |
|---|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 215.6 | 8066.8 | 1.33 | 162.6 W |
| NVIDIA B300 | 196.5 | 3963.7 | 0.65 | 303.2 W |
| NVIDIA H200 | 188.6 | 5888.9 | 1.09 | 173.5 W |
| NVIDIA H100 80GB HBM3 | 186.1 | 5907.8 | 1.1 | 168.6 W |
| NVIDIA B200 | 183.8 | 6194 | 0.59 | 309.1 W |
| NVIDIA L40S | 157.9 | 6663.6 | 1.14 | 137.9 W |
| NVIDIA A100 80GB SXM4 | 124.1 | 3280.4 | 0.85 | 145.9 W |
| NVIDIA A100 40GB SXM4 | 119.1 | 3248.6 | 1.21 | 98.4 W |
| NVIDIA A10G | 103.2 | 2267.8 | 1.03 | 100.5 W |
| NVIDIA L4 | 75.03 | 2086 | 1.44 | 52.0 W |
What the numbers show. Across 10 GPUs measured on our own bench, RTX PRO 6000 Blackwell Workstation Edition is fastest at 216 tok/s. The slowest, L4, manages 75.0, so the spread is 2.9x from top to bottom. L4 is the most efficient, 75.0 tok/s at 52W. Per dollar of launch price, A10G gives the most (36.9 tok/s per $1,000).
About GLM-4.7-Flash. GLM-4.7-Flash: from zai-org, 31B parameters, on Hugging Face since January 2026, MIT licence. 2,136,434 downloads in the last 30 days and 2 community quantizations.
How it compares. H100 80GB HBM3: GLM-4.7-Flash 186.1 tok/s, Gemma 4 31B 76.59 (31B), Nemotron 3.5 Lightning 30B A3B 323.6 (32B), Nemotron-3-Nano-30B-A3B 316.5 (32B), Qwen3 30B A3B 283.8 (31B). 3 of 4 beat GLM-4.7-Flash here. Nemotron 3.5 Lightning 30B A3B, Nemotron-3-Nano-30B-A3B and Qwen3 30B A3B are mixture-of-experts, so per token they compute only a slice of their size.
Cost on a rented GPU. 1M generated tokens of GLM-4.7-Flash: $1.10 on a A100 40GB SXM4 ($0.47/hr, 2.3 hours), $1.39 on a RTX PRO 6000 Blackwell Workstation Edition ($1.08/hr, 77 min, 1.3x the cost).
GLM-4.7-Flash: cost per 1M generated tokens on rented GPUs
| GPU | Cheapest rate | Speed (tok/s) | Cost per 1M generated tokens |
|---|---|---|---|
| NVIDIA A100 40GB SXM4 | $0.47/hr | 119.1 | $1.10 |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | $1.08/hr | 215.6 | $1.39 |
| NVIDIA L40S | $0.79/hr | 157.9 | $1.39 |
| NVIDIA L4 | $0.44/hr | 75.03 | $1.63 |
| NVIDIA A100 80GB SXM4 | $0.95/hr | 124.1 | $2.12 |
| NVIDIA H100 80GB HBM3 | $2.14/hr | 186.1 | $3.19 |
| NVIDIA H200 | $3.59/hr | 188.6 | $5.29 |
| NVIDIA B200 | $5.98/hr | 183.8 | $9.04 |
| NVIDIA B300 | $6.94/hr | 196.5 | $9.81 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for GLM-4.7-Flash. 30+ tok/s: 10 (RTX PRO 6000 Blackwell Workstation Edition, B300, H200). 30 tok/s is roughly where replies outpace reading.
Reading your prompt. Before GLM-4.7-Flash writes anything it reads the input: 8066.8 tok/s on the RTX PRO 6000 Blackwell Workstation Edition (0.5s for a 4,000-token prompt), 2086.0 on the L4 (1.9s). Long documents and big code files feel this number more than the generation speed.
VRAM for GLM-4.7-Flash. Measured peak 17.9GB, so 24GB is the smallest common card size; smallest card it ran on: A10G (24GB). With long context: Q4_K_M 23GB (tested), Q2_K 15GB, Q3_K_M 19GB, Q5_K_M 27GB, Q6_K 32GB.
Power on GLM-4.7-Flash. Most efficient: L4, 52W, 0.19 kWh per 1M generated tokens. Hungriest: B200, 309W, 0.47 kWh. At $0.15/kWh: $0.029 per 1M generated tokens.
Our verdict
Fastest on GLM-4.7-Flash: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 215.6 tok/s. Cheapest to rent per job: NVIDIA A100 40GB SXM4, $1.10 per 1M generated tokens.
FAQ
- What GPU do I need to run GLM-4.7-Flash?
- About 19GB. Smallest card that ran it: NVIDIA A10G (24GB).
- How much does it cost to run GLM-4.7-Flash in the cloud?
- $1.10 per 1M generated tokens on a NVIDIA A100 40GB SXM4 at $0.47/hr, cheapest of 9 rentable cards we measured.
- Can I run GLM-4.7-Flash on a 12GB, 16GB or 24GB card?
- It used 17.9GB at the precision we tested. 12GB: no; 16GB: only at Q2_K (15GB); 24GB: yes.
- Is the H100 80GB HBM3 or the A100 80GB SXM4 faster for GLM-4.7-Flash?
- The H100 80GB HBM3: 186.1 vs 124.1 tok/s, 50% faster on our bench.
How we test
llama.cpp llama-bench at Q4_K_M, 512-token prompt and 128 generated tokens, three runs after a warmup, full GPU offload, with power and VRAM sampled throughout. Token generation is memory-bandwidth-bound, so the ranking tracks bandwidth closely, which makes it a fair guide to cards we haven't run yet.