GLM-5.1

GLM-5.1 is an open-weight large language model developed by the Chinese AI company Zhipu AI, which markets its products internationally under the brand Z.ai. Released on April 7, 2026, with weights published the following day, it is the company's flagship model for agentic engineering and the direct successor to GLM-5.^[1][2][3] GLM-5.1 is a mixture-of-experts (MoE) model that Z.ai labels 744B-A40B, about 744 billion total parameters with about 40 billion active per token, distributed on Hugging Face under the permissive MIT License.^[19][2] Its headline feature is long-horizon autonomy: Z.ai says the model can work continuously on a single coding task for up to eight hours, refining its own strategy across hundreds of iterations rather than answering in one shot.^[1][5]

Overview

GLM-5.1 is a text-in, text-out model positioned squarely at software engineering and tool-driven agent work.^[1][6] Z.ai describes it as a refinement of the GLM-5 series tuned for "long-horizon" execution, the ability to keep making forward progress on a hard problem long after a typical model has, in the company's words, run out of ideas.^[5] Where earlier releases were built for minute-scale interactions, GLM-5.1 is designed to plan, run experiments, read the results, find the blockers, and try again over many rounds without a human stepping in.^[1][5]

The model reaches state-of-the-art results among open models on several agentic and coding benchmarks, and Z.ai claims it edges past closed frontier systems including GPT-5.1-Codex-Max's lineage, Claude Opus 4.6, and Gemini 3 Pro on the SWE-Bench Pro software-engineering benchmark.^[1][3][7] It ships with a 200,000-token context window and supports up to 128,000 output tokens in a single response.^[6][8]

Developer: Zhipu AI / Z.ai and the GLM line

Zhipu AI is a Beijing-based AI company that spun out of Tsinghua University and develops the GLM (General Language Model) family.^[3][9] The company uses the consumer and developer brand Z.ai for its international platform and API, while operating the BigModel.cn platform in China.^[3][7] Zhipu has been on the United States Entity List since January 2025, which restricts its access to American-made AI accelerators, a constraint that shaped how the GLM-5 series was trained.^[10][9]

The GLM line had already moved through several generations before GLM-5.1. The wiki documents GLM-4.5 and GLM-4.6 from the 4.x series, the multimodal GLM-4-Voice, and the GLM-5 base model. GLM-5.1 is not a separate architecture from GLM-5 so much as the next iteration of it: a model in the same MoE family, retrained and post-trained to push much harder on coding, tool use, and sustained autonomous execution.^[1][5][2]

Release

Z.ai announced GLM-5.1 on April 7, 2026, and open-sourced the weights on Hugging Face on April 8, 2026, alongside an API rollout.^[1][2][11] Regional and English-language press carried the launch over April 7 through April 9, 2026, with most outlets dating the announcement to April 7 or 8.^[3][7][12] The release was framed by the company as its "most intelligent flagship model to date" and, in Z.ai's own marketing, the strongest open-source model then available.^[12][7]

The launch coincided with a price increase for Zhipu's cloud API. The company raised pricing on model calls by roughly 8 to 17 percent relative to its earlier GLM-5 Turbo offering, a move several outlets noted ran against the broader industry trend of falling token prices.^[11][12] Coverage tied the pricing and release to Zhipu's business position: the company had reported a widening loss for 2025 and had raised prices repeatedly through early 2026, and its shares moved sharply on launch-week disclosures.^[11]

Variants and sizes

At launch GLM-5.1 was distributed as a single flagship model rather than a family of differently sized checkpoints. The base weights are published in BF16, and a large ecosystem of quantized builds (for runtimes such as llama.cpp, Ollama, and LM Studio) appeared on Hugging Face shortly after release.^[2][4]

Scroll sideways for more →

Z.ai also exposed the model through its GLM Coding Plan tiers (Lite, Pro, and Max) and made it compatible with agentic coding clients including Claude Code, Cline, and other OpenAI-compatible tools.^[1][13]

Architecture and training

GLM-5.1 uses the architecture Z.ai labels glm_moe_dsa, a mixture-of-experts design combined with a sparse-attention mechanism the company derived from research on efficient long-context attention.^[2][20] The published config.json specifies 78 transformer layers, of which the first three carry dense feed-forward blocks and the remaining 75 carry MoE blocks; each MoE block holds 256 routed experts plus one shared expert, with exactly 8 routed experts activated per token. Hidden size is 6,144, there are 64 attention heads, and the vocabulary is 154,880 tokens.^[20] The attention stack is Multi-head Latent Attention combined with DeepSeek Sparse Attention, whose lightweight indexer picks a top-k subset of 2,048 key positions per query instead of attending over the whole sequence; that is what keeps the 200,000-token context window tractable. It is not a linear-attention design, and the config carries no linear-attention or sliding-window component.^[20]

The 744B and 754B figures

Z.ai's own model index in the zai-org/GLM-5 repository labels GLM-5.1, in both its BF16 and FP8 forms, 744B-A40B, the same designation it gives GLM-5 and GLM-5.2.^[19] The GLM-5.1 model card and Z.ai's GLM-5.1 documentation page publish no parameter count at all, so a "754B" figure attributed to either of them is misattributed.^[2][6] That number comes from aggregators and deployment guides reading the safetensors metadata of the published checkpoint, which totals 753,864,139,008 parameters and rounds to 754 billion.^[4][14] The gap is the multi-token-prediction draft block that config.json's num_nextn_predict_layers: 1 setting adds for speculative decoding: summing the shapes in that block gives 9,952,920,576 parameters, and subtracting it leaves 743,911,218,432, which is the 744 billion Z.ai publishes.^[20] The GLM-5.1 checkpoint is byte-for-byte the same size as GLM-5's, 59,870 tensors and 1.51 TB on disk, so GLM-5.1 is not a larger model than GLM-5 on any basis.^[20] Artificial Analysis, which maintains its own model records, lists GLM-5.1 at 744B total and 40B active.^[21]

The GLM-5 series is notable for the hardware it is reported to have run on. Press and community write-ups said the GLM-5 line, including GLM-5.1, was trained without Nvidia accelerators, on a large cluster of Huawei Ascend 910-series chips together with Huawei's MindSpore framework, and put that cluster among the largest non-Nvidia training deployments disclosed to date, which several outlets framed as evidence that competitive frontier models can be built outside the Nvidia ecosystem under export-control pressure.^[10][15] Z.ai itself did not disclose this. The GLM-5 technical report, which is the paper the GLM-5.1 model card asks readers to cite, never names MindSpore, gives no chip count, and does not describe the training hardware; its Chinese-silicon section is about deployment, reporting full-stack inference adaptation across seven domestic platforms (Huawei Ascend, Moore Threads, Hygon, Cambricon, Kunlunxin, MetaX and Enflame).^[19][2] For context, the GLM-5 base model scaled the family from 355 billion parameters (32 billion active) up to 744 billion (40 billion active) and was trained on about 28.5 trillion tokens of pre-training data, integrating a sparse-attention mechanism for the first time.^[16] GLM-5.1 carries that lineage forward with a post-training and reinforcement learning recipe aimed at long-horizon agency, using what Z.ai describes as asynchronous RL infrastructure.^[1][5]

Capabilities

The defining capability of GLM-5.1 is sustained autonomous execution. Rather than producing a one-shot answer, the model runs an experiment, analyzes, and optimizes loop, repeatedly reviewing its own strategy, recognizing dead ends, and trying new approaches.^[1][5] Z.ai reports demonstrations in which the model worked for more than eight hours on a single task, including building a complete Linux desktop environment from scratch through hundreds of autonomous plan-execute-analyze-optimize iterations in one uninterrupted session.^[1][5] In other demonstrations it iterated on performance tuning over hundreds of rounds and thousands of tool calls, for example pushing a CUDA kernel from a roughly 2.6x to a far larger speedup and improving a vector-database workload through hundreds of iterations.^[5][6]

On coding and agentic AI workflows, GLM-5.1 is optimized for clients such as Claude Code and supports function calling, structured tool use, and OpenAI-SDK-compatible access.^[6][7] Beyond software engineering, the model posts strong scores on general reasoning, mathematics, and web-browsing agent benchmarks, indicating it is a broadly capable system rather than a coding-only specialist.^[2][4]

Benchmarks

The figures below come from the benchmark table Z.ai published on the GLM-5.1 model card in April 2026. Scores are self-reported by Z.ai unless otherwise noted, and benchmark results for any model should be read with that caveat.^[2][4][1]

Scroll sideways for more →

Z.ai summarized the coding results by saying that across the combined average of SWE-Bench Pro, Terminal-Bench, and NL2Repo, GLM-5.1 ranked third globally, first among Chinese models, and first among open-source models.^[12][2] Reviewers measuring it against Claude Opus 4.6 reported coding performance in the low-to-mid-90s percent range relative to that model, with Z.ai's CEO stating that coding capability "approaches" Claude Opus 4.6.^[11][17]

License and availability

Scroll sideways for more →

The permissive MIT license, rather than a custom community license, was repeatedly highlighted as a draw: it allows download, modification, fine-tuning, and commercial deployment without usage restrictions or fees, subject to the practical reality that self-hosting a 744B model whose BF16 checkpoint runs to 1.51 TB requires substantial hardware.^[2][17][20]

Reception

Coverage of GLM-5.1 was broadly positive about its open-weight, low-cost challenge to closed frontier labs. Outlets framed it as the first open-source model to lead the SWE-Bench Pro leaderboard, and several described it as the strongest open-weight model available at the time of release.^[1][3][7] Technical writers emphasized two themes in particular: the long-horizon autonomy, which they treated as a meaningful step beyond minute-scale chat interactions, and the training story, since the model reached frontier-adjacent results on Huawei Ascend hardware with no Nvidia chips, an outcome read as significant given U.S. export controls.^[5][10][15]

Constellation Research and others noted that GLM-5.1 topped public leaderboards on Hugging Face and chat arenas around launch, while some reviewers tempered the enthusiasm by stressing that the most striking claims came from Z.ai's own benchmarks and demonstrations.^[3][1] Independent aggregate scoring is more measured. On the Artificial Analysis Intelligence Index v4.1, as measured on July 27, 2026, GLM-5.1 scores 40.2, essentially level with GLM-5's 39.5 and well behind the 51.1 of the later GLM-5.2; index scores from earlier versions of that benchmark are not comparable with v4.1 figures.^[21] The simultaneous price increase drew comment as an unusual move in a market trending toward cheaper tokens, and was generally read against Zhipu's financial position rather than as a sign of weakness in the model.^[11][12]

Limitations

Several limitations and caveats apply. Many of the most dramatic capability claims, the eight-hour autonomous runs and the leaderboard-topping benchmark scores, are self-reported by Z.ai or shown in company demonstrations, and independent reproduction of the long-horizon results at the scale Z.ai describes was limited at launch.^[1][5] There is also no GLM-5.1 technical report: the model card points readers to the GLM-5 paper, so the architecture figures above are read off the shipped config.json and checkpoint metadata rather than from a datasheet written for this release, and Z.ai has published no training-token count, compute budget, or data mix specific to GLM-5.1.^[2][20]

Practical access is constrained by scale: at 744 billion parameters the full model is out of reach for typical local hardware, so most users rely on the hosted API or quantized community builds, and the MIT license does not change those hardware requirements.^[2][17] The model is text-only, without the image or audio input found in some rival systems, and its knowledge and benchmark coverage reflect an early-2026 snapshot.^[6][2] Finally, as with any frontier coding agent, sustained autonomous operation raises the usual concerns about unsupervised tool use, which the long-horizon framing makes more salient rather than less.^[1][5]

References

1. ^MarkTechPost, "Z.AI Introduces GLM-5.1: An Open-Weight 754B Agentic Model That Achieves SOTA on SWE-Bench Pro and Sustains 8-Hour Autonomous Execution," marktechpost.com/...ns-8-hour-autonomous-execution
2. ^Hugging Face, "zai-org/GLM-5.1" model card, huggingface.co/...GLM-5.1
3. ^Constellation Research, "Z.ai ups ante in open-source LLMs with GLM-5.1," constellationr.com/...ante-open-source-llms-glm-51
4. ^LLM-Stats, "GLM-5.1 Benchmarks, Pricing & Context Window," llm-stats.com/...glm-5.1
5. ^The Decoder, "Zhipu AI's GLM-5.1 can rethink its own coding strategy across hundreds of iterations," the-decoder.com/...y-across-hundreds-of-iterations
6. ^Z.ai Developer Documentation, "GLM-5.1 - Overview," docs.z.ai/...glm-5.1
7. ^TestingCatalog, "Zhipu AI launches open-source GLM-5.1 model for coding tasks," testingcatalog.com/...m-5-1-model-for-coding-tasks
8. ^APXML, "GLM-5.1: Specifications and GPU VRAM Requirements," apxml.com/...glm-51
9. ^AIMadeTools, "What is Z.ai (Zhipu)? The Lab Behind GLM-5.1," aimadetools.com/...what-is-z-ai-zhipu
10. ^Awesome Agents, "GLM-5.1 Tops SWE-Bench Pro With Zero NVIDIA Hardware," awesomeagents.ai/...5-1-swe-bench-pro-huawei-chips
11. ^Yahoo Finance / Bloomberg, "Zhipu Raises AI Model Prices 8%-17% As GLM-5.1 Launches," finance.yahoo.com/...ses-ai-model-prices-131652529
12. ^Futunn News, "The release of Zhipu GLM-5.1: A 10% price increase against market trends, open-source models surpass closed-source counterparts," news.futunn.com/...ase-of-zhipu-glm-5-1-a-10-price
13. ^Atlas Cloud Blog, "LLM GLM-5.1 is coming soon to Atlas Cloud: Opus 4.6-Level Coding with 8-Hour Autonomous Execution," atlascloud.ai/...-with-8-hour-autonomous-execution
14. ^Spheron Blog, "Deploy GLM-5.1 on GPU Cloud: Self-Host the 754B MoE Model (2026 Guide)," spheron.network/...deploy-glm-5-1-gpu-cloud
15. ^Let's Data Science, "How China's GLM-5 Works: 744B Model on Huawei Chips," letsdatascience.com/...-model-glm-5-without-nvidia
16. ^Z.ai Developer Documentation, "GLM-5 - Overview," docs.z.ai/...glm-5
17. ^MindStudio, "GLM 5.1: The Open-Source Model That Matches GPT and Claude on Coding," mindstudio.ai/...glm-5-1-open-source-coding-model
18. ^Together AI, "GLM-5.1 API," together.ai/...glm-51
19. ^zai-org, "GLM-5" GitHub repository (model index and technical report links), github.com/...GLM-5
20. ^zai-org, "GLM-5.1 config.json and model.safetensors.index.json," Hugging Face, accessed July 27, 2026, huggingface.co/...config.json
21. ^Artificial Analysis, "GLM-5.1 model page," accessed July 27, 2026, artificialanalysis.ai/...glm-5-1
22. ^Hugging Face, "zai-org/GLM-5" model card, huggingface.co/...GLM-5