SELF-HOSTING # Qwen3.8 27B: Self-Hosting Estimate Qwen3.8 27B (Alibaba (Qwen)) on different GPUs: configuration, estimated performance and use cases. Data updated: 2026-09-16 [← All self-hosting estimates](https://www.tandiu.cn/en/local-deploy) Total parameters**28B**Sets weight VRAM Active parameters**28B**Sets decode compute and speed License**apache-2.0**[Official repo →](https://huggingface.co/Qwen/Qwen3.8-27B) Public benchmarks**AA 20.1 · LiveBench 60.5** Vendor API price**$0.50 / $3.00**input / output per 1M tokens · [Token prices →](https://www.tandiu.cn/en/token-market) Vendor API output speed**—**measured by Artificial Analysis; compare with local estimates below ## B300 288GB 288 GB · 8,000 GB/s · Top-tier Blackwell card (no official channels to China) Suggested precision **FP16/BF16 (full precision)** GPUs needed **1priced as 1 × 8-GPU HGX node** Est. decode speed **144 tok/s**single-stream theoretical ceiling Est. batched throughput **out 245 tok/s**in 5,261 tok/s · many concurrent requests Hardware list price **$299,058–$498,429**Cloud rental (live): $7.88–$7.88/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 **1** Est. decode speed **144 tok/s**single-stream theoretical ceiling Est. batched throughput **out 245 tok/s**in 5,845 tok/s · many concurrent requests Hardware list price **$29,943–$44,915** Hardware market price **$89,757–$179,514** 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 **1priced as 1 × 8-GPU HGX node** Est. decode speed **144 tok/s**single-stream theoretical ceiling Est. batched throughput **out 245 tok/s**in 5,261 tok/s · many concurrent requests Hardware list price **$299,058–$498,429**Cloud rental (live): $7.45–$12.51/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 **1priced as 1 × 8-GPU HGX node** Est. decode speed **86 tok/s**single-stream theoretical ceiling Est. batched throughput **out 147 tok/s**in 2,312 tok/s · many concurrent requests Hardware list price **$279,837–$499,708**Cloud rental (live): $2.99–$4.61/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 **FP16/BF16 (full precision)** GPUs needed **1priced as 1 × 8-GPU HGX node** Est. decode speed **60 tok/s**single-stream theoretical ceiling Est. batched throughput **out 102 tok/s**in 2,312 tok/s · many concurrent requests Hardware list price **$239,860–$831,514**Cloud rental (live): $1.93–$4.11/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 **FP16/BF16 (full precision)** GPUs needed **1priced as 1 × 8-GPU HGX node** Est. decode speed **60 tok/s**single-stream theoretical ceiling Est. batched throughput **out 102 tok/s**in 2,312 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 **FP16/BF16 (full precision)** GPUs needed **1** Est. decode speed **72 tok/s**single-stream theoretical ceiling Est. batched throughput **out 122 tok/s**in 346 tok/s · many concurrent requests Hardware list price **n/a** Hardware market price **$11,968–$16,455** 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 **FP16/BF16 (full precision)** GPUs needed **1** Est. decode speed **35 tok/s**single-stream theoretical ceiling Est. batched throughput **out 59 tok/s**in 729 tok/s · many concurrent requests Hardware list price **$14,960–$15,558** Hardware market price **$16,156–$17,951** 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 **1** Est. decode speed **58 tok/s**single-stream theoretical ceiling Est. batched throughput **out 98 tok/s**in 1,758 tok/s · many concurrent requests Hardware list price **n/a** Hardware market price **$11,968–$29,919** 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 **FP16/BF16 (full precision)** GPUs needed **2** Est. decode speed **14 tok/s**single-stream theoretical ceiling Est. batched throughput **out 24 tok/s**in 1,496 tok/s · many concurrent requests Hardware list price **n/a** Hardware market price **$32,911–$35,903** 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 **FP16/BF16 (full precision)** GPUs needed **1** Est. decode speed **32 tok/s**single-stream theoretical ceiling Est. batched throughput **out 55 tok/s**in 1,169 tok/s · many concurrent requests Hardware list price **$16,025**Cloud rental (live): $1.20–$2.27/hr (2026-11-05) Hardware market price **$17,586** 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 **FP16/BF16 (full precision)** GPUs needed **1** Est. decode speed **25 tok/s**single-stream theoretical ceiling Est. batched throughput **out 43 tok/s**in 1,057 tok/s · many concurrent requests Hardware list price **$11,393** Hardware market price **$12,716–$13,464** 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 **FP16/BF16 (full precision)** GPUs needed **2** Est. decode speed **48 tok/s**single-stream theoretical ceiling Est. batched throughput **out 49 tok/s**in 724 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **n/a**Cloud rental (live) ×2: $1.12–$2.14/hr (2026-11-05) Hardware market price **$16,156** 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 **FP16/BF16 (full precision)** GPUs needed **2** Est. decode speed **31 tok/s**single-stream theoretical ceiling Est. batched throughput **out 32 tok/s**in 1,016 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **$13,464–$19,447**Cloud rental (live) ×2: $1.20–$1.82/hr (2026-11-05) Hardware market price **$13,464–$19,447** Typical use SME teams building inference services, medium-low concurrency ## L20 48GB 48 GB · 864 GB/s · Compliant inference card Suggested precision **FP16/BF16 (full precision)** GPUs needed **2** Est. decode speed **31 tok/s**single-stream theoretical ceiling Est. batched throughput **out 32 tok/s**in 335 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **$6,985–$7,025** Hardware market price **$10,472** 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 **FP16/BF16 (full precision)** GPUs needed **4** Est. decode speed **129 tok/s**single-stream theoretical ceiling Est. batched throughput **out 131 tok/s**in 2,351 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **n/a**Cloud rental (live) ×4: $1.77–$2.13/hr (2026-11-05) Hardware market price **$26,927** 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 **FP16/BF16 (full precision)** GPUs needed **4** Est. decode speed **97 tok/s**single-stream theoretical ceiling Est. batched throughput **out 99 tok/s**in 1,667 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **$9,873**Cloud rental (live) ×4: $1.77–$2.12/hr (2026-11-05) Hardware market price **$16,755–$23,935** 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 **FP16/BF16 (full precision)** GPUs needed **4** Est. decode speed **73 tok/s**single-stream theoretical ceiling Est. batched throughput **out 74 tok/s**in 1,650 tok/s · many concurrent requests · PCIe multi-GPU penalty Hardware list price **$7,778–$11,369**Cloud rental (live) ×4: $1.35–$1.74/hr (2026-11-05) Hardware market price **$15,558–$17,951** 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.