GLM-5.3-Flash: Self-Hosting Estimate
GLM-5.3-Flash (Zhipu AI) on different GPUs: configuration, estimated performance and use cases.
Data updated: 2026-09-16B300 288GB
288 GB · 8,000 GB/s · Top-tier Blackwell card (no official channels to China)- Suggested precision
- FP16/BF16 (full precision)
- GPUs needed
- 4priced as 1 × 8-GPU HGX node
- Est. decode speed
- 889 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,511 tok/sin 32,500 tok/s · many concurrent requests
- Hardware list price
- $299,058–$498,429Cloud rental (live) ×4: $31.54–$31.54/hr (2026-11-04)
- Hardware market price
- $1,047,167
- Typical use
- Enterprise-grade ultra-large-scale training and high-concurrency inference clusters
GB200 192GB (single card equivalent)
192 GB · 8,000 GB/s · Top-tier Blackwell card (no official channels to China)- Suggested precision
- FP16/BF16 (full precision)
- GPUs needed
- 8
- Est. decode speed
- 1,778 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 3,022 tok/sin 72,222 tok/s · many concurrent requests
- Hardware list price
- $239,544–$359,316
- Hardware market price
- $718,058–$1,436,115
- Typical use
- Ultra-large-scale trillion-parameter model training and inference clusters (networked with Grace CPU)
B200 180GB
180 GB · 8,000 GB/s · Top-tier Blackwell card (no official channels to China)- Suggested precision
- FP16/BF16 (full precision)
- GPUs needed
- 8priced as 1 × 8-GPU HGX node
- Est. decode speed
- 1,778 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 3,022 tok/sin 65,000 tok/s · many concurrent requests
- Hardware list price
- $299,058–$498,429Cloud rental (live) ×8: $59.58–$100.08/hr (2026-11-05)
- Hardware market price
- $448,786–$523,584
- Typical use
- Enterprise-grade large-scale online inference and training
H200 141GB
141 GB · 4,800 GB/s · Top-tier export-controlled card (channels to China are volatile)- Suggested precision
- FP16/BF16 (full precision)
- GPUs needed
- 8priced as 1 × 8-GPU HGX node
- Est. decode speed
- 1,067 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,813 tok/sin 28,571 tok/s · many concurrent requests
- Hardware list price
- $279,837–$499,708Cloud rental (live) ×8: $23.93–$36.86/hr (2026-11-05)
- Hardware market price
- $224,393–$233,369
- Typical use
- Enterprise-grade production environment high-concurrency inference
H100 SXM5 80GB
80 GB · 3,350 GB/s · Top-tier export-controlled card (no official channels to China)- Suggested precision
- INT8
- GPUs needed
- 8priced as 1 × 8-GPU HGX node
- Est. decode speed
- 1,489 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 2,531 tok/sin 28,571 tok/s · many concurrent requests
- Hardware list price
- $239,860–$831,514Cloud rental (live) ×8: $15.47–$32.90/hr (2026-11-05)
- Hardware market price
- $320,732
- Typical use
- Enterprise-grade production environment high-concurrency inference/training
H800 SXM 80GB
80 GB · 3,350 GB/s · Top-tier export-controlled card (discontinued)- Suggested precision
- INT8
- GPUs needed
- 8priced as 1 × 8-GPU HGX node
- Est. decode speed
- 1,489 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 2,531 tok/sin 28,571 tok/s · many concurrent requests
- Hardware list price
- $293,207–$311,158
- Hardware market price
- $403,907–$418,867
- Typical use
- Enterprise-grade production deployment (original China-compliant model, now discontinued)
H20 96GB
96 GB · 4,000 GB/s · Export-compliant special supply card (significantly reduced compute power)- Suggested precision
- INT8
- GPUs needed
- 8
- Est. decode speed
- 1,778 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 3,022 tok/sin 4,276 tok/s · many concurrent requests
- Hardware list price
- n/a
- Hardware market price
- $95,741–$131,644
- Typical use
- Suboptimal choice in export-compliant environments, medium-to-large scale inference deployment (FP16 compute power is only about 15% of H100, but higher memory bandwidth offers an advantage in decode-intensive scenarios)
A800 PCIe 80GB
80 GB · 1,935 GB/s · Compliant special supply card (export compliance status adjusted multiple times)- Suggested precision
- INT8
- GPUs needed
- 8
- Est. decode speed
- 860 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 877 tok/sin 5,408 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $119,676–$124,463
- Hardware market price
- $129,250–$143,612
- Typical use
- Medium-scale inference/training deployment in export-compliant environments
Ascend 910C 128GB
128 GB · 3,200 GB/s · Huawei dual-chip data center card (domestically produced)- Suggested precision
- FP16/BF16 (full precision)
- GPUs needed
- 8
- Est. decode speed
- 711 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,209 tok/sin 21,724 tok/s · many concurrent requests
- Hardware list price
- n/a
- Hardware market price
- $95,741–$239,353
- Typical use
- Domestically produced large model training and inference clusters, CloudMatrix and other super-node solutions
Ascend 910B 64GB
64 GB · 400 GB/s · Domestic alternative (not subject to US export controls)- Suggested precision
- INT8
- GPUs needed
- 8
- Est. decode speed
- 178 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 302 tok/sin 9,244 tok/s · many concurrent requests
- Hardware list price
- n/a
- Hardware market price
- $131,644–$143,612
- Typical use
- Domestic alternative, not subject to US export controls, medium-to-large scale inference/training deployment, memory bandwidth significantly lower than NVIDIA cards of similar VRAM capacity
RTX PRO 6000 Blackwell 96GB
96 GB · 1,792 GB/s · Professional workstation/server card (China-specific version is 6000D 84GB)- Suggested precision
- INT8
- GPUs needed
- 8
- Est. decode speed
- 796 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 812 tok/sin 8,667 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $128,200Cloud rental (live) ×8: $9.62–$18.15/hr (2026-11-05)
- Hardware market price
- $140,691
- Typical use
- Single-card large VRAM inference, local deployment of medium MoE models, private all-in-one machine
RTX PRO 5500 Blackwell 84GB
84 GB · 1,398 GB/s · Professional workstation card (specs are NVIDIA preliminary data)- Suggested precision
- INT8
- GPUs needed
- 8
- Est. decode speed
- 621 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 634 tok/sin 7,835 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $91,147
- Hardware market price
- $101,725–$107,709
- Typical use
- Workstation large VRAM inference, departmental private deployment
RTX PRO 5000 Blackwell 48GB
48 GB · 1,344 GB/s · Professional workstation card- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 8
- Est. decode speed
- 1,195 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,219 tok/sin 4,472 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- n/aCloud rental (live) ×8: $4.49–$8.55/hr (2026-11-05)
- Hardware market price
- $64,625
- Typical use
- Small to medium model inference, development testing, single-machine multi-card private deployment
L40S 48GB
48 GB · 864 GB/s · Compliant inference card- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 8
- Est. decode speed
- 768 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 783 tok/sin 6,275 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $53,854–$77,790Cloud rental (live) ×8: $4.81–$7.30/hr (2026-11-05)
- Hardware market price
- $53,854–$77,790
- Typical use
- SME teams building inference services, medium-low concurrency
L20 48GB
48 GB · 864 GB/s · Compliant inference card- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 8
- Est. decode speed
- 768 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 783 tok/sin 2,071 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $27,941–$28,101
- Hardware market price
- $41,887
- Typical use
- SME teams building inference services, medium-low concurrency
RTX 5090 32GB
32 GB · 1,792 GB/s · Consumer flagship (not sold in China, 5090D for domestic market)- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 8
- Est. decode speed
- 1,593 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,625 tok/sin 7,263 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- n/aCloud rental (live) ×8: $3.54–$4.26/hr (2026-11-05)
- Hardware market price
- $53,853
- Typical use
- Local inference and development for individuals/small teams, multi-card VRAM pooling for quantized MoE
RTX 5090D 24GB
24 GB · 1,344 GB/s · Consumer-grade/small-scale self-built- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 16 (exceeds one node: multi-node needed)
- Est. decode speed
- 2,389 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 2,437 tok/sin 10,296 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $39,491Cloud rental (live) ×16: $7.07–$8.48/hr (2026-11-05)
- Hardware market price
- $67,019–$95,741
- Typical use
- Personal development/small-scale verification/POC, not recommended for production-level high concurrency
RTX 4090D 24GB
24 GB · 1,008 GB/s · Consumer-grade/small-scale self-built- Suggested precision
- INT4 (AWQ/GPTQ)
- GPUs needed
- 16 (exceeds one node: multi-node needed)
- Est. decode speed
- 1,792 tok/ssingle-stream theoretical ceiling
- Est. batched throughput
- out 1,828 tok/sin 10,192 tok/s · many concurrent requests · PCIe multi-GPU penalty
- Hardware list price
- $31,113–$45,475Cloud rental (live) ×16: $5.41–$6.98/hr (2026-11-05)
- Hardware market price
- $62,232–$71,806
- Typical use
- Personal development/small-scale verification/POC, not recommended for production-level high concurrency
FAQ
FAQ
How is the estimated decode speed calculated?
Estimated speed = (memory bandwidth × GPU count) ÷ (active parameters × bytes per parameter): the memory-bound upper limit for one request (roofline). It scales linearly with GPU count and ignores interconnect overhead, batching and framework losses, so real results are usually well below it.
Can I trust the speed numbers for consumer GPUs?
The linear assumption is reasonable for NVLink data-center cards (H/B series) but too optimistic for consumer PCIe cards without NVLink (RTX 5090D/4090D); multi-GPU consumer setups lose far more to communication, so equal-looking numbers do not mean equal performance.
How is “Suggested precision” chosen?
It is the highest precision that fits within 8 GPUs; beyond that it is flagged as multi-node.
Where do the prices come from?
USD converted at ¥6.68 per USD. Cloud rental is the international vast.ai market rate, not a China purchase price.