GLM-5.2Open Weight
Released
GLM-5.2 is an AI model released by Z.ai on Tuesday, Jun 16 2026, 70 days after GLM-5.1. It is an open-weight model — the trained weights are available to download and run. It is a 744B parameter model with a 1M token context window. Benchmark results (shown below) cover BullshitBench v2, ProgramBench, SWE-Bench Pro, DeepSWE 1.1, Next.js Evals, Frontier-Bench v0.1, and 4 more.
Get Z.ai releases and AI news.
Every other Monday
API pricing
- InputWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it.
- $1.40
- Cached inputA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again.
- $0.26
- OutputWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one.
- $4.40
Available from
| Decart | — | — | $0.2925 | $1.80 | 1M | mxfp4 |
|---|---|---|---|---|---|---|
| DeepInfra | — | — | $0.5625 | $1.80 | 1M | fp4 |
| StreamLake | — | — | $0.6426 | $2.0196 | 1M | fp8 |
| Novita | — | — | $0.6496 | $2.0416 | 1M | fp8 |
| DigitalOcean | — | — | $0.70 | $2.20 | 1M | — |
| SiliconFlow | — | — | $0.70 | $2.20 | 1M | fp8 |
| CoreWeave | — | — | $0.76 | $2.42 | 1M | fp4 |
| AtlasCloud | — | — | $0.938 | $2.948 | 1M | fp8 |
| Phala | — | — | $1.26 | $3.00 | 1M | fp8 |
| Alibaba | — | — | $0.966 | $3.036 | 1M | fp8 |
| Morph | — | — | $0.093 | $3.553 | 1M | fp8 |
| Inceptron | — | — | $1.39 | $4.39 | 1M | fp4 |
| InferenceNet | — | — | $0.20 | $4.40 | 1M | fp4 |
| Cloudflare | — | — | $1.18 | $4.40 | 256K | — |
| Baidu | — | — | $1.40 | $4.40 | 1M | fp8 |
| BaseTen | — | — | $1.40 | $4.40 | 1M | fp8 |
| Friendli | — | — | $1.40 | $4.40 | 1M | — |
| GMICloud | — | — | $1.40 | $4.40 | 1M | fp8 |
| Mistral | — | zdr | $1.40 | $4.40 | 1M | nvfp4 |
| Nebius | — | — | $1.40 | $4.40 | 1M | fp4 |
| Parasail | — | — | $1.40 | $4.40 | 256K | fp4 |
| Together | — | — | $1.40 | $4.40 | 1M | — |
| Venice | — | — | $1.40 | $4.40 | 1M | fp8 |
| Z.AI | — | — | $1.40 | $4.40 | 1M | fp8 |
| Mistral | eu | — | $1.54 | $4.84 | 1M | — |
| BaseTen | — | fast | $2.10 | $6.60 | 1M | fp8 |
| Relace | — | — | $0.171 | $7.20 | 1M | — |
| Alibaba | — | fast | $2.31 | $7.26 | 1M | fp8 |
| Baidu | — | fast | $2.25 | $7.88 | 1M | fp4 |
| Wafer | — | — | $0.19 | $8.00 | 1M | — |
| Decart | — | fast | $2.50 | $9.00 | 1M | fp4 |
Benchmarks
Coding
ProgramBenchProgram reconstruction — The AI receives a working program and its documentation, then builds a replacement from scratch without the original source code, internet access or decompilation. The score is the percentage of 200 programs that pass every behavioral test. We record each model's best published mini-SWE-agent result, including higher reasoning efforts where available. Partial test-pass rates and almost-solved programs do not count toward this score. Equal scores share a rank here; the official board also uses partial progress to break ties. Higher is better.via ProgramBench
0%
#6 of 20Best: Claude Opus 5 · 4.5%
SWE-Bench ProAgentic coding — Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better.
62.1%
#10 of 23Best: Claude Fable 5.1 · 81.2%
DeepSWE 1.1Agentic coding — Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better.
44%
#26 of 32Best: Gemini 4 Argon · 77.9%
Next.js EvalsNext.js coding — Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better.
81%
#9 of 32Best: Claude Fable 5.1 · 97%
Terminal & CLI
Frontier-Bench v0.1Agentic computer work — A hard, ever-evolving set of real computer tasks — coding, system administration, data work, and more — that an AI agent has to complete on its own. Run by the Harbor / Laude Institute team as the successor to Terminal-Bench (v0.1 is the first release of the task set). The score is the share of tasks solved. Higher is better.
5.1%
#9 of 9Best: Claude Opus 5 · 43.3%
Terminal-Bench 2.1Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better.
81%
#18 of 32Best: DeepSeek-V4.1-Flash · 90.6%
Reasoning & science
Humanity's Last ExamMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better.via BenchLM
40.5%no tools
54.7%with tools
no tools
#10 of 22Best: Claude Fable 5.1 · 60.9%
with tools
#17 of 45Best: Claude Opus 5.5 · 67.7%
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
91.2%
#13 of 59Best: GPT-6 Astra · 96%
Knowledge work
GDPval-AA v2Knowledge work — economically valuable knowledge work (v2, re-based Elo)
1514
#14 of 20Best: Claude Opus 5 · 1861
Robustness
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.
31%
#53 of 84Best: Claude Opus 4.8 · 95%
Source: ProgramBench, retrieved 11 September 2026 · Source: BenchLM, retrieved 11 July 2026. Other sources are identified on the linked benchmark pages.
Compare GLM-5.2 with
Frequently asked questions
GLM-5.2 was released by Z.ai on Tuesday, Jun 16 2026.