Qwen 3.8 Flash Next 176B-A6B — B300 vs H200
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on Qwen 3.8 Flash Next 176B-A6B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
B300 hits 81505 tok/s/chip for $0.01 per million tokens at 71 tok/s/user on Qwen 3.8 Flash Next 176B-A6B. No H200 data at this operating point.
B300: 69672 tok/s/chip, $0.01 per million tokens at 104 tok/s/user on Qwen 3.8 Flash Next 176B-A6B. H200 is unmeasured here.
At 137 tok/s/user on Qwen 3.8 Flash Next 176B-A6B, B300 delivers 52685 tok/s/chip at $0.01 per million tokens; H200 hasn't been benchmarked at this target. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:81505.3H200:— | B300:69672.2H200:— | B300:52685.2H200:— |
| Cost ($/M tok) | B300:$0.008H200:— | B300:$0.009H200:— | B300:$0.012H200:— |
| tok/s/MW | B300:42897518H200:— | B300:36669601H200:— | B300:27729063H200:— |
| Concurrency | B300:~16H200:— | B300:~13H200:— | B300:~10H200:— |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.