DeepSeek V4 Pro 1.6T
8K/1K
Hyperscaler cost ↓ Lower is better Source: InferenceX & SemiAnalysis Market July 2026 AI Cloud TCO Model
| Model · Scenario | B200 | MI355X | B300 | GB200 NVL72 | GB300 NVL72 |
|---|---|---|---|---|---|
DeepSeek V4 Pro 1.6T8K/1K | TRTLLM · FP4 | TRTLLM · FP4 | Dynamo vLLM · FP4 | ||
DeepSeek V4 Pro 1.6TLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | vLLM · FP4 | MoRI SGLang · FP4 | vLLM · FP4 | — no data for this scenario | Dynamo vLLM · FP4 |
Kimi K3 2.8TLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | Dynamo vLLM · FP4 · STP | — no data for this scenario | vLLM · FP4 | Dynamo vLLM · FP4 | — no data for this scenario |
MiniMax M3 428B8K/1K | ATOM¹ · FP4 | ||||
MiniMax M3 428BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | — no data for this scenario | vLLM · FP4 | — no data for this scenario | — no data for this scenario |
GLM5.2Long Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | — no data for this scenario | — no exact @50 result | — no data for this scenario | — no data for this scenario |
Qwen3.5 397B8K/1K | TRTLLM · FP4 | Dynamo SGLang · FP8 · STP | |||
Qwen3.5 397BLong Context Multi-Turn Realistic Agentic Scenario (AgentX) | — no exact @50 result | — no data for this scenario | SGLang · FP4 | — no data for this scenario | Dynamo SGLang · FP4 |
DeepSeek R1 0528 671BMaintenance8K/1K | TRTLLM · FP4 | ATOM¹ · FP4 | Dynamo TRTLLM · FP4 | Dynamo TRTLLM · FP4 · STP | Dynamo TRTLLM · FP8 · STP |
Kimi K2.5/2.6/2.7-Code 1TDeprecated8K/1K | Dynamo TRTLLM · FP4 · STP | ATOM¹ · FP4 · STP | vLLM · FP4 · STP | Dynamo TRTLLM · FP4 · STP | Dynamo TRTLLM · FP4 · STP |
GLM5/5.1 744BDeprecated8K/1K | SGLang · FP4 | SGLang · FP8 | SGLang · FP4 | Dynamo TRTLLM · FP4 · STP | |
gpt-oss 120BDeprecated8K/1K | TRTLLM · FP4 · STP | ATOM¹ · FP4 · STP | — no data for this scenario | — no exact @50 result | — no data for this scenario |
MiniMax M2.5/2.7 230BDeprecated8K/1K | vLLM · FP4 · STP | ATOM¹ · FP4 · STP | vLLM · FP4 · STP | Dynamo vLLM · FP4 · STP | Dynamo vLLM · FP4 · STP |
Llama 3.3 70B InstructDeprecated8K/1K | TRTLLM · FP4 · STP | vLLM · FP4 · STP | — no data for this scenario | — no data for this scenario | — no data for this scenario |
8K/1K
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
8K/1K
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
8K/1K
Long Context Multi-Turn Realistic Agentic Scenario (AgentX)
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
8K/1K
Current cost and change versus the latest validated platform result 30–60 days earlier.
Platforms without a valid 30-day comparison show current cost only.
If a chip does not have FP4 spec decoding available, the next best available configuration is used.