Model details
GLM-5.1 is Z.AI's next-generation flagship model for agentic engineering, positioned to keep working effectively over long, multi-hour sessions rather than plateauing after early gains. It is trained with multi-turn supervised fine-tuning, reinforcement learning, and a process-quality evaluation framework aimed at improving stability, consistency, and tool use across extended tasks, with the goal of running autonomously for up to eight hours in a single pass. Compared with its predecessor GLM-5, it shows stronger engineering intelligence across autonomous planning, sustained execution, bug fixing, and strategy iteration.
In practice, the model is aimed at complex software engineering and repository-level development workflows. Its release benchmarks report a state-of-the-art score of 58.4 on SWE-Bench Pro for complex software engineering tasks, and a wide lead over GLM-5 on NL2Repo repo generation and Terminal-Bench 2.0 real-world terminal tasks. Weights are published openly under an MIT-style distribution on Hugging Face at zai-org/GLM-5.1, with code mirrored at github.com/zai-org/GLM-5, making GLM-5.1 a natural fit for teams building coding agents, deep debuggers, and other long-horizon developer tools that need durable judgment and iteration rather than one-shot answers.
Transparent token rates
Rates are shown per one million tokens. Combined means one million input plus one million output tokens.
Z.AI
2026 04 07 · Research GLM 5.1: Towards Long Horizon Tasks GLM 5.1 is our next generation flagship model for agentic engineering, with significantly stronger coding capabilities than its predecessor. It achieves state of the art performance on SWE Bench Pro and leads GLM 5 by a wide margin on NL2Repo (repo generation) a
Z.AI
Z.AI Releases GLM-5.1: A Next-Generation Agentic Model Built for Long-Horizon Engineering Tasks and Rewrites CUDA Kernels
Z.AI
Z.ai raises prices for its most advanced AI model, GLM-5.1, by at least 8% compared to GLM-5 Turbo, joining Alibaba and Tencent as demand for agentic AI surges. This signals vendor-level price pressure for advanced models and could increase deployment and inference costs for developers and enterprises.
Z.AI
Z.ai releases GLM-5.1, a 754B-parameter model (1.51TB on Hugging Face) distributed under an MIT license; it matches the parameter count of GLM-5 and is presented as aimed at long-horizon tasks.
Z.AI
According to a recent LinkedIn post from GMI Cloud, the company is highlighting what it describes as Day 0 support for the newly released GLM-5.1 model from Z.ai. T...
Merge Gateway
InferenceX provides a technical deep-dive covering GLM-5 and GLM-5.1, noting that GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active) with 28.5T pre-training tokens, and that GLM-5.1 is a point release on the same architecture with stronger coding and SOTA SWE-Bench Pro. The article cites the GLM-5.1's distinguishing claim per the coverage is long-horizon durability: sustaining optimization over hundreds of rounds and thousands of tool calls where earlier models exhaust their repertoire early. The article notes GLM-5 integrates DeepSeek Sparse Attention (DSA) to reduce deployment cost while preserving long-
Merge Gateway
Z.ai's official release-notes index explicitly lists GLM-5.1 under the 2026-04-07 entry, describing it as designed for long-horizon tasks with the ability to work independently for up to 8 hours in a single run, covering the full loop from planning and execution to iterative refinement and final delivery. The entry not The release notes confirm GLM-5.1 was built with multi-turn SFT, RL, and process-based training methods, and serves as the documented baseline against which later releases GLM-5.2 (2026-06-16) and GLM-5.3 (2026-08-18) are compared in subsequent entries. This provides authoritative dating and positioning of GLM-5.1 with
Z.AI
OpenRouter's GLM 5.1 listing gives developers concrete, current deployment data: release date April 7, 2026, 205K context, and a published list price of $0.9086/$2.856 per 1M tokens (a 35% discount shown alongside the standard $1.40/$4.40 Z.ai rate). The page confirms the model's core positioning—"can work independentl The same listing provides a developer-actionable snapshot of 16 hosting providers with input/output/cache-read pricing, P50 latency, throughput (tok/s), and uptime—e.g., Chutes at $0.98/$3.08 with 35 tok/s, DeepInfra at $1.05/$3.50, Crusoe with 0.55s latency at 40 tok/s, and Friendli at 0.64s/65 tok/s. Average throughp
