Z.ai Launches GLM 5.2: Open-Source 744B Model With 1M-Token Context Window

Z.ai released GLM 5.2 on June 13, 2026, a 744-billion-parameter Mixture-of-Experts model featuring a 1-million-token context window and an MIT open-source license with no regional restrictions. The model currently ranks fourth among 124 models on BenchLM's provisional leaderboard and is the top-ranked open-weight model across three major long-horizon coding benchmarks, placing alongside proprietary frontier models. A new architectural feature called IndexShare reduces per-token compute by 2.9x at long context lengths, while an improved multi-token-prediction layer boosts speculative-decoding acceptance by around 20%. Priced at roughly one-sixth the cost of leading frontier models, GLM 5.2 is positioned as a cost-effective option, though experts warn that its large context window can drive up API costs significantly if developers send unnecessarily large prompts. Teams are advised to track input token usage per request and send only the minimum context required, rather than defaulting to filling the full 1-million-token window on every call.