Today we’re announcing Not Diamond Code, the world’s most powerful intelligent model router for long-horizon coding agents.
Not Diamond works with any gateway or harness, including Claude Code, to select the best model and reasoning effort for each step, reducing costs by 20-65% without impacting quality.
Tomas Hernando Kofman on X: "Today we’re announcing Not Diamond Code, the world’s most powerful intelligent model router for long-horizon coding agents. Not Diamond works with any gateway or harness, including Claude Code, to select the best model and reasoning effort for each step, reducing costs by 20-65% https://t.co/io9biZOxbM"
Not Diamond Code achieves Pareto-optimal performance on SWE-PolyBench-Verified and LongCodeQA, approximating the quality of Opus 4.8 Xhigh at 39–61% lower cost. Evaluations were run with Claude Code as the harness routing between Haiku 4.5, Sonnet 4.6, and Opus 4.8 at variousThe greatest opportunity for routing however is in leveraging multiple providers across closed-source and open-source models. On SWE-PolyBench-Verified, routing between Anthropic models and GLM 5.2 and DeepSeek V4 Flash improved performance by 3.6% and increased savings from 39%Not Diamond Code is based on a novel routing technique that makes recommendations by predicting future rewards and costs for a given model and reasoning effort at each step of an agent sequence. Recommendations are made in a cache-aware manner that optimizes for long-trajectoryBefore each turn, Not Diamond considers current and prior session states, message and token counts, task complexity, implicit developer feedback signals, and KV cache state to predict the future cost and reward of each model and reasoning-effort recommendation. It then selectsNot Diamond is the first harness- and gateway-agnostic model router. It works natively with Claude Code without requiring any change in behavior to save you inference spend automatically.We founded Not Diamond to build the routing infrastructure for the multi-model future, all the way back to releasing the world’s first open source intelligent model router in December 2023. A world with many diverse models is more performant and cost-efficient than one in whichWe are selectively onboarding teams to Not Diamond Code early access. Run it on your own coding-agent workflows and compare routing against your current default-model strategy. Learn more and request access here:
