## Latest OpenRouter Models That Are GLM 5.2 Competitors (June 2026) **GLM 5.2** (Z.ai, June 16, 2026) is the #1 open-weight AI model on the Artificial Analysis Intelligence Index at **51**. It's a 753B-param / 40B-active MoE with a 1M context window. ### 🥇 Direct Open-Weight Competitors on OpenRouter | Model | Lab | Key Spec | Intelligence Index | Context | OpenRouter Price (input/output per M tokens) | |---|---|---|---|---|---| | **GLM 5.2** | Z.ai | 753B/40B active MoE | **51** (#1 open) | 1M | ~$0.45 / $3.31 | | **DeepSeek V4 Pro** | DeepSeek | 1.6T/~200B active MoE | 44 | 128-200K | ~$0.55-0.75 / $1.50-2.00 | | **DeepSeek V4 Flash** | DeepSeek | 284B/13B active MoE | — | 1M | **$0.14/$0.28** (cheapest) | | **NVIDIA Nemotron 3 Ultra** | NVIDIA | — | **48** (#2 open) | — | Varies | | **Kimi K2.7 Code** | Moonshot AI | 1T/30B active MoE | ~43 | 256K | ~$0.95 / $4.00 | | **Kimi K2.6** | Moonshot AI | — | 43 | — | Varies | | **MiniMax M3** | MiniMax | Multimodal, long-context | 44 | — | ~$0.30/M input | | **Qwen 3.7** | Alibaba | Frontier open-weight | — | — | Varies | | **Mistral (latest)** | Mistral AI | European open-weight | — | — | Varies | ### When GLM 5.2 Wins - **Long-horizon agentic coding** — 1M context, top Intelligence Index score (51) - **Front-end coding** — #2 on Code Arena WebDev (second only to Claude Fable 5) - **Self-host cost efficiency** — Only 40B active params, fits on fewer GPUs ### When Competitors Win - **DeepSeek V4 Flash** → Cost-sensitive bulk coding ($0.14/M input, 150x cheaper than GPT-5.5) - **DeepSeek V4 Pro** → Max SWE-bench score (80.6%), ecosystem maturity - **Kimi K2.7 Code** → MCP tool-heavy agents, vision input (MoonViT), first-party Kimi Code workspace - **NVIDIA Nemotron 3 Ultra** → Best U.S. open-weight option (48 on AI Index) ### The Routing Strategy (June 2026 Recommendation) Most production teams are using **OpenRouter to route across all three**: GLM 5.2 for long-horizon coding, DeepSeek V4 Flash/Pro for cost-sensitive SWE, and Kimi K2.7 Code for MCP-heavy agent workloads. *(Sources: OpenRouter Blog June 27, 2026; Andrew.ooo comparison June 19, 2026; OpenRouter API model listing)*