Running Qwen3.8-Flash-Next on H200
Quick answer
Qwen3.8-Flash-Next sustains 16,353 tokens/s per GPU on H200 at 50 tokens/s per user, which works out to $0.021 per million tokens at hyperscaler pricing, served by SGLang. Fastest measured TTFT: 0.7 ms; fastest TPOT: 0.0 ms (each the best across all configs, not one run).
Benchmarked configs
5
Serving engines
SGLang
Precisions
fp8
Run dates
2026-08-27 → 2026-08-27
Throughput at every interactivity target
Serving is a trade-off: push more concurrent users through a GPU and each user's tokens arrive slower. The ladder below reads the measured frontier at each per-user speed target on the AgentX agentic coding workload, using the best engine and precision at that point.
| Per-user target | Tokens/s per GPU | $ / 1M tokens | Engine | Precision |
|---|---|---|---|---|
| 30 tok/s | 16,803 | $0.020 | SGLang | fp8 |
| 50 tok/s | 16,353 | $0.021 | SGLang | fp8 |
| 75 tok/s | 15,416 | $0.022 | SGLang | fp8 |
| 100 tok/s | 13,709 | $0.025 | SGLang | fp8 |
| 150 tok/s | - | - | SGLang | fp8 |
| 200 tok/s | - | - | SGLang | fp8 |
What serving actually costs
Converting the 50 tokens/s per user operating point to $ per million total tokens across rental pricing tiers from the SemiAnalysis AI Cloud TCO model.
| Pricing tier | $ / GPU / hr | $ / 1M tokens |
|---|---|---|
| Hyperscaler | $1.22 | $0.021 |
| Neocloud | $1.59 | $0.027 |
| Retail | $2.05 | $0.035 |
Frequently asked questions
- How fast is Qwen3.8-Flash-Next on H200?
- At an interactivity target of 50 tokens/s per user on the AgentX agentic coding workload, H200 sustains 16,353 tokens/s per GPU serving Qwen3.8-Flash-Next with SGLang in FP8. Peak measured throughput across all configs is 17,081 tokens/s per GPU.
- How much does it cost to serve Qwen3.8-Flash-Next on H200?
- $0.021 per million total tokens at hyperscaler $/GPU/hr pricing, at 50 tokens/s per user. Neocloud and retail rental tiers are tabulated above; slower interactivity targets lower the cost further.
- Which serving engines run Qwen3.8-Flash-Next on H200?
- The runs behind this page used SGLang in FP8. Engines are rebuilt and re-benchmarked continuously, so the best config can change between visits.
- How are these Qwen3.8-Flash-Next numbers measured?
- Every number is measured on real H200 hardware by the InferenceX fleet, sweeping concurrency on the AgentX agentic coding workload to trace the throughput-versus-interactivity frontier; the newest run landed on 2026-08-27. The same derivation powers the InferenceX overview leaderboard.