SemiAnalysis Breaks Down 1.3B Tokens/$ AgentX Record: vLLM + Inferact Disaggregated Serving Beats Aggregated B300 by 4x, While Cline Highlights Multi-Harness RL
SemiAnalysis published the architectural breakdown behind the 1.3B tokens per dollar TCO record on AgentX: vLLM and Inferact utilized Prefill/Decode disaggregated serving at hyperscaler volume pricing to achieve 4x better cost-efficiency than an aggregated NVIDIA B300 configuration. Concurrently, open-source coding agent Cline reported that DeepSeek-V4.1-Flash-1017 beats GPT-5.6-Sol on Terminal-Bench at ~17x lower cost (+35% over prior V4.1-Flash) purely by scaling RL training across multiple coding agent harnesses.
Key Takeaways
- SemiAnalysis confirms vLLM and Inferact achieve 1.3B tokens per dollar TCO on AgentX via disaggregated serving—4x better than an aggregated NVIDIA B300 configuration
- Cline reports that DeepSeek-V4.1-Flash-1017 improves prior V4.1-Flash Terminal-Bench performance by 35%, beating GPT-5.6-Sol at ~17x lower cost
- The 35% agentic coding leap was achieved without base algorithmic changes by scaling reinforcement learning directly across multiple coding agent harnesses
Linked Models & Agents:
