name = "GLM-4.6"
description = "Late GLM-4 workhorse for coding agents, reasoning, and structured tasks"
family = "glm"
release_date = "2025-09-30"
last_updated = "2025-09-30"
attachment = false
reasoning = true
temperature = true
tool_call = true
knowledge = "2025-04"
open_weights = true

[limit]
context = 204_800
output = 131_072

[modalities]
input = ["text"]
output = ["text"]

[[weights]]
label = "Hugging Face"
url = "https://huggingface.co/zai-org/GLM-4.6"

[[benchmarks]]
name = "Artificial Analysis Coding Index"
score = 29.5
metric = "index"
source = "https://openrouter.ai/z-ai/glm-4.6/benchmarks"
date = "2026-05-22"

[[benchmarks]]
name = "SciCode"
score = 38.4
metric = "percent correct"
source = "https://openrouter.ai/z-ai/glm-4.6/benchmarks"
date = "2026-05-22"

[[benchmarks]]
name = "Terminal-Bench Hard"
score = 25
metric = "success rate"
source = "https://openrouter.ai/z-ai/glm-4.6/benchmarks"
date = "2026-05-22"

[[benchmarks]]
name = "SWE-Bench Pro"
score = 9.67
metric = "resolve rate"
dataset = "public"
source = "https://labs.scale.com/leaderboard/swe_bench_pro_public"