GLM Latest

GLM Latest functions as a compatibility alias that always resolves to Zhipu's newest thinking-oriented GLM release, which is currently GLM 5.3 Thinking. The alias pattern lets developers keep a single model identifier while Zhipu rotates underlying versions, avoiding hardcoded migrations as newer generations land. Because the alias explicitly targets the "thinking" variant, it is designed for workflows that benefit from deliberate, multi-step reasoning rather than instant chat responses, and reasoning effort cannot be disabled since the lowest setting is mapped to maximum thinking.

In practice, GLM Latest fits agentic and research-heavy applications where long contexts and structured reasoning matter. Reported benchmark scores highlight an Intelligence Index of 52.6, graduate-level GPQA Diamond reasoning at 89.5%, Humanity's Last Exam at 41.1%, and strong instruction-following and long-context reasoning results, alongside solid agentic coding performance on Terminal-Bench. A roughly one-million-token context window with around 131K tokens of output headroom makes it well suited for document analysis, tool-using agents, and multi-turn workflows that combine retrieval, planning, and code generation.

OpenRouter~z-ai/glm-latestglm

Open provider page Provider docs

Quick Info

Provider: OpenRouter
Model key: ~z-ai/glm-latest
Release date: Aug 19, 2026
Last updated: Aug 19, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.063
Output token cost: $6.30

Limits

Output tokens: 131,072 tokens
Context window: 1,048,576 tokens

Transparent token rates

Compare GLM Latest 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 Latest

OpenRouter

Coverage

The Price Per Token news index entry dated 8/19/2026 lists "Z.ai: GLM Latest (z-ai/glm-latest)" with the description "This model always redirects to the latest GLM model from Z.ai," confirming the alias's semantic behavior on OpenRouter. The index also tracks related GLM-family entries on OpenRouter, including z-ai/glm While the entry does not specify which GLM version the alias currently points to, it corroborates the August 19 addition date and the always-redirects-to-latest semantics, giving developers a clean reference for when the GLM Latest route became available on OpenRouter and how it sits alongside the explicit versioned GL

OpenRouter

CoverageBenchmark

Veso Research's ranking matrix independently lists GLM-5.3, GLM-5.3-BF16, and GLM-5.3-Flash as distinct entries, providing a second source that corroborates the existence of a Flash variant in addition to the main model. The GLM-5.3 row shows 753.3B parameters, an open-weights MoE architecture, and a release date of 20 The aggregator's release-date stamps disagree with Z.ai's 18 August 2026 date reported by AZ Labs, and its 1.3M context figure for Flash differs from the 1M context described in the AZ Labs piece and the Z.ai release notes. SWE-Bench, Terminal-Bench, and other evaluation columns are N/A for the GLM-5.3 entries, so this

OpenRouter

Coverage

AZ Labs reports that on August 19, 2026, OpenRouter added the z-ai/glm-latest alias at 17:19:39 SAST, describing it as a paid alias whose route update targeted GLM-5.3 rather than constituting a separate release from Z.ai. The article documents Z.ai's official GLM-5.3 spec — text model with a 1M-token context, forced r The piece frames these as a single provider-availability wave rather than four independent launches, and emphasizes that Z.ai has not announced a separate Flash launch. For OpenRouter users of the GLM Latest alias, the practical takeaway is that the route was redirected toward GLM-5.3 on August 19, with the cheaper Fla

OpenRouter

Official sourceBenchmark

OpenRouter's model page for `z-ai/glm-latest` is the authoritative primary source confirming that the alias always redirects to Z.ai's latest GLM model, and explicitly identifies the current target as "Latest GLM 5.3." The page documents an OpenAI-compatible POST endpoint at `https://openrouter.ai/api/v1/chat/completio The page surfaces OpenRouter-specific optional headers (`HTTP-Referer`, `X-Title`) that allow apps to appear on the leaderboards and points developers to the `frameworks` documentation for third-party SDKs. While pricing, context window, and capability tags were not captured in the supplied excerpt, the model page is t