GLM-5.2 (CanopyWave)

The supplied research evidence for GLM-5.2 (CanopyWave) is limited to a single 404 page-not-found response from llmpricing.dev, and no creator-side documentation such as release notes, a model card, or a technical report was returned. Because no source excerpt establishes the model's creator, training methodology, architecture, parameter count, or benchmark performance, no substantive narrative about GLM-5.2's design or lineage can be supported in this overview. In practical terms, there is no verifiable information to describe GLM-5.2's intended use cases, qualitative strengths, or ideal fit for tasks such as reasoning, tool calling, or long-context work. Readers seeking confirmed details about this model should consult the original creator's official channels once those sources become available, since the current evidence base does not support any further descriptive claims.

Without corroborating technical documentation, any forward-looking commentary about GLM-5.2's advancements relative to earlier generations or competing systems would be speculative. Consequently, this overview cannot responsibly characterize the model as a step forward in efficiency, reasoning quality, context handling, or open-weight accessibility, even though the catalog indicates that open weights are available. A meaningful comparison will only be possible once independent sources document its training data composition, evaluation results, and deployment characteristics.

LLM Gatewaycanopywave/glm-5.2glm

Open provider page Provider docs

Quick Info

Provider: LLM Gateway
Model key: canopywave/glm-5.2
Release date: Jun 13, 2026
Last updated: Jun 13, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.40
Output token cost: $4.40

Limits

Output tokens: 32,768 tokens
Context window: 200,000 tokens

Transparent token rates

Compare GLM-5.2 (CanopyWave) pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about GLM-5.2 (CanopyWave)

LLM Gateway

Coverage

GLM-5.2 was released on June 13, 2026 by Z.ai (formerly Zhipu AI) as an open-weights large language model. At launch it was the top open-weights model on the Artificial Analysis Intelligence Index, ranking first among open models and fourth overall, though Moonshot AI's Kimi K3 (July 27, 2026) has since taken the open- Against frontier closed models, GLM-5.2 still beats GPT-5.5 on coding benchmarks while trailing the current closed frontier, and Z.ai has followed up with GLM-5.3 (August 14, 2026) which improves on coding results but has not yet released public weights. The article, authored by PromptQuorum's founder, frames GLM-5.2 a

LLM Gateway

CoverageBenchmark

BenchmarkList's compilation aggregates GLM-5.2 scores across 111 evaluations, attributing the model to Z.ai. Notable agentic results include a final account value of $8,313.78 on Vending-Bench 2 (rank 5/54, 92nd percentile), 34.6% Pass@1 on Tau3-Banking (rank 27/174, 85th percentile, with GLM-5.3 leading at 50.3%), 82. The page includes a pricing row ($0.94 input / $3.00 output per 1M tokens) that is serving-side rate data and therefore out of scope for model-focused news. The comparison matrix also references other models (Kimi K3, Qwen3.8 variants, GLM-5.3, Claude Fable 5.1, Claude Opus 5) for context. All evidence relates to Z.ai'

LLM Gateway

CoverageBenchmark

Layer3 Labs published a benchmark guide (reviewed by Jonathan West, updated September 7, 2026) that separates independently verified GLM-5.2 results from Zhipu's self-reported numbers. The strongest independent signal is a first-place finish on Design Arena's Code Categories leaderboard, a blind human-preference test w The guide flags specific point scores on SWE-bench Pro, Terminal-Bench 2.1, and FrontierSWE as coming from Zhipu's own technical report and model card, noting they are directionally corroborated by independent evaluators but not reproduced exactly under Zhipu's precise methodology. The headline "near-parity with Claude