Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on MiniMax M3 428B. 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 →
B200 / GB300 NVL72 on MiniMax M3 428B at 159 tok/s/user: 34188 / 32568 tok/s/chip, $0.01 / $0.02 per million tokens. B200 is 40% cheaper per token; B200 delivers 5% more tok/s/chip.
Around the middle of the 100–336 tok/s/user interactivity band, at 218 tok/s/user on MiniMax M3 428B: B200 runs 20797 tok/s/chip at $0.02/M tokens, GB300 NVL72 runs 21832 at $0.03/M. B200 is 27% cheaper per token; GB300 NVL72 delivers 5% more tok/s/chip.
Setting 277 tok/s/user as the target on MiniMax M3 428B, B200 produces 9798 tok/s/chip ($0.05 per million tokens) and GB300 NVL72 produces 13103 ($0.05). Cost per token is essentially tied; GB300 NVL72 delivers 34% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:34188.4GB300 NVL72:32567.6 | B200:20797.3GB300 NVL72:21832.5 | B200:9798.3GB300 NVL72:13103.4 |
| Cost ($/M tok) | B200:$0.014GB300 NVL72:$0.020 | B200:$0.023GB300 NVL72:$0.029 | B200:$0.049GB300 NVL72:$0.049 |
| tok/s/MW | B200:19993207GB300 NVL72:15362054 | B200:12162187GB300 NVL72:10298334 | B200:5730024GB300 NVL72:6180856 |
| Concurrency | B200:~20GB300 NVL72:~29 | B200:~14GB300 NVL72:~20 | B200:~6GB300 NVL72:~24 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
MiniMax M3 428B 8K / 1K Total Tokens per $1 TCO vs. Interactivity