DeepSeekv4 Pro 0813 1.6T · Chip comparison

DeepSeekv4 Pro 0813 1.6T — B300 vs GB300 NVL72

Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on DeepSeekv4 Pro 0813 1.6T. 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 / GB300 NVL72 on DeepSeekv4 Pro 0813 1.6T at 71 tok/s/user: 15153 / 84710 tok/s/chip, $0.04 / $0.01 per million tokens. GB300 NVL72 is 447% cheaper per token; GB300 NVL72 delivers 459% more tok/s/chip.

Around the middle of the 34–181 tok/s/user interactivity band, at 108 tok/s/user on DeepSeekv4 Pro 0813 1.6T: B300 runs 9081 tok/s/chip at $0.07/M tokens, GB300 NVL72 runs 6610 at $0.10/M. B300 is 40% cheaper per token; B300 delivers 37% more tok/s/chip.

Setting 145 tok/s/user as the target on DeepSeekv4 Pro 0813 1.6T, B300 produces 5439 tok/s/chip ($0.12 per million tokens) and GB300 NVL72 produces 3647 ($0.18). B300 is 52% cheaper per token; B300 delivers 49% 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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/chip)
B300:15153.1GB300 NVL72:84709.9
B300:9080.6GB300 NVL72:6609.8
B300:5438.8GB300 NVL72:3647.0
Cost ($/M tok)
B300:$0.041GB300 NVL72:$0.008
B300:$0.069GB300 NVL72:$0.097
B300:$0.115GB300 NVL72:$0.176
tok/s/MW
B300:7975298GB300 NVL72:39957505
B300:4779271GB300 NVL72:3117845
B300:2862507GB300 NVL72:1720278
Concurrency
B300:~12GB300 NVL72:~1020
B300:~7GB300 NVL72:~5
B300:~3GB300 NVL72:~2

Inference Performance

Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.

Total Tokens per $1 USD (Owning - Hyperscaler) vs. Interactivity

DeepSeek V4 Pro 1.6T FP4 8K / 1K Source: SemiAnalysis InferenceX™

TCO $/chip/hr: H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68

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.

No data available

Please change the model, sequence, precision, date range or chip selection.

Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip