Hyperscaler cost ↓ Lower is better Source: InferenceX & SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
| Model · Scenario | B200 · Reference | MI355X | B300 | GB200 NVL72 | GB300 NVL72 |
|---|---|---|---|---|---|
DeepSeek V4 Pro 0813 1.6TLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | vLLM · FP4 | $0.03130% more expensive than B200 SGLang · FP4 | $0.01922% cheaper than B200 vLLM · FP4 | $0.04694% more expensive than B200 llm-d vLLM · FP4 | $0.01152% cheaper than B200 Dynamo vLLM · FP4 |
Kimi K3 2.8TLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | Dynamo vLLM · FP4 | $0.08420% cheaper than B200 vLLM · FP4 | $0.103About the same cost as B200 vLLM · FP4 | $0.1137% more expensive than B200 Dynamo vLLM · FP4 | $0.08024% cheaper than B200 Dynamo vLLM · FP4 |
MiniMax M3 428BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | $0.014No B200 baseline to compare against vLLM · FP4 | — no exact @50 result | — no exact @50 result | — no exact @50 result |
GLM5.2/GLM5.3 744BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | — no exact @50 result | — no exact @50 result | — no exact @50 result | — no exact @50 result |
Qwen3.5 397BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | $0.008No B200 baseline to compare against SGLang · FP4 | $0.014No B200 baseline to compare against SGLang · FP8 | — no exact @50 result | — no exact @50 result |
Qwen3.8 Flash Next 176BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no data for this scenario | — no data for this scenario | — no exact @50 result | — no data for this scenario | — no data for this scenario |
DeepSeek V4 Pro 0813 1.6T8K/1K | TRTLLM · FP4 | $0.15943% more expensive than B200 ATOM¹ · FP4 | $0.13420% more expensive than B200 TRTLLM · FP4 | $0.514360% more expensive than B200 Dynamo vLLM · FP4 | $0.21391% more expensive than B200 Dynamo SGLang · FP4 · STP |
Qwen3.5 397B8K/1K | TRTLLM · FP4 | $0.05656% more expensive than B200 SGLang · FP4 | $0.04011% more expensive than B200 SGLang · FP4 | $0.142295% more expensive than B200 Dynamo SGLang · FP8 · STP | $0.04011% more expensive than B200 Dynamo TRTLLM · FP4 · STP |
DeepSeek R1 0528 671BMaintenanceModel is no longer actively benchmarked.8K/1K | TRTLLM · FP4 | $0.157105% more expensive than B200 ATOM¹ · FP4 | $0.13981% more expensive than B200 Dynamo TRTLLM · FP4 | $0.192149% more expensive than B200 Dynamo SGLang · FP8 · STP | $0.230200% more expensive than B200 Dynamo TRTLLM · FP8 · STP |
Kimi K2.5/2.6/2.7-Code 1TDeprecatedModel is no longer actively benchmarked.8K/1K | Dynamo TRTLLM · FP4 · STP | $0.13257% more expensive than B200 ATOM¹ · FP4 · STP | $0.179112% more expensive than B200 vLLM · FP4 · STP | $0.0797% cheaper than B200 Dynamo TRTLLM · FP4 · STP | $0.09817% more expensive than B200 Dynamo TRTLLM · FP4 · STP |
GLM5/5.1 744BDeprecatedModel is no longer actively benchmarked.8K/1K | SGLang · FP4 | $0.431150% more expensive than B200 SGLang · FP8 | $0.21022% more expensive than B200 SGLang · FP4 | $0.26051% more expensive than B200 Dynamo TRTLLM · FP4 · STP | $0.28163% more expensive than B200 Dynamo TRTLLM · FP4 · STP |
gpt-oss 120BDeprecatedModel is no longer actively benchmarked.8K/1K | TRTLLM · FP4 · STP | $0.010About the same cost as B200 ATOM¹ · FP4 · STP | — no data for this scenario | — no exact @50 result | — no data for this scenario |
MiniMax M2.5/2.7 230BDeprecatedModel is no longer actively benchmarked.8K/1K | vLLM · FP4 · STP | $0.05123% more expensive than B200 ATOM¹ · FP4 · STP | $0.05124% more expensive than B200 vLLM · FP4 · STP | $0.041About the same cost as B200 Dynamo vLLM · FP4 · STP | $0.04713% more expensive than B200 Dynamo vLLM · FP4 · STP |
Llama 3.3 70B InstructDeprecatedModel is no longer actively benchmarked.8K/1K | TRTLLM · FP4 · STP | $0.145124% more expensive than B200 vLLM · FP4 · STP | — no data for this scenario | — no data for this scenario | — no data for this scenario |
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
— = no result. ∞ = B200 baseline unavailable.
If a chip does not have FP4 spec decoding available, the next best available configuration is used.
DeepSeek V4 Pro 0813 1.6T
Kimi K3 2.8T
MiniMax M3 428B
GLM5.2/GLM5.3 744B
Qwen3.5 397B
gpt-oss 120B
Llama 3.3 70B Instruct