On this page

OpenRouter listed Z.ai: GLM 5.3 FlashX Legal on 2026-11-02. The models API stamps created 1793626154 → 2026-11-02 13:29:14 UTC, 21:29:14 Beijing time (Asia/Shanghai). The formal call id is z-ai/glm-5.3-flashx-legal; the canonical slug is z-ai/glm-5.3-flashx-legal-20261102. The OpenRouter model page also uses the short name GLM-Legal.

Below: the formal id, live price rows, published context and completion figures, and how Z.ai’s docs place the legal-domain variant. We haven't run prompts, contract fixtures, or agent harnesses against the live route.

Related reading: OpenCode makes GLM 5.3 Flash free for a limited time, Claude for Legal Adds Ready-to-Use Artifacts Gallery, OpenRouter Adds GPT-6.1 Terra Pro for Higher-Quality Reasoning, and Cloudflare Fixed Cross-Request KV Cache Exposure in Workers AI GLM-5.3.

Formal id and listing facts

FieldValue
OpenRouter idz-ai/glm-5.3-flashx-legal
Canonical slugz-ai/glm-5.3-flashx-legal-20261102
Page nameZ.ai: GLM 5.3 FlashX Legal
Released (page)Nov 2, 2026
API created2026-11-02 13:29:14 UTC ＝ 21:29:14 Asia/Shanghai
Modalitytext + image + video → text
Context1,048,576 (page chip also 1.0M)

OpenRouter describes GLM-5.3-FlashX-Legal as Z.ai’s legal-domain variant of GLM-5.3-FlashX: multimodal input, legal-domain mid-training and post-training with reinforcement learning in task environments modeled on real legal workflows, aimed at document understanding, legal reasoning, and multi-step work such as legal research, contract review, cross-document consistency checks, and drafting or revising legal documents. It keeps FlashX’s 1M context, mandatory thinking, function calling, structured output, and streaming, and is framed for agent setups wired to legal data and document-editing tools (MCP services). Those lines are vendor copy—not DigIntoAI measurements of legal quality or drafting accuracy.

Live prices — cache lines now match

OpenRouter’s provider table (Z.ai as the sole listed provider):

Billable line (OpenRouter)Per 1M tokens
Input$0.37
Output$1.25
Cache read$0.075

Z.ai’s Text Models pricing row for GLM-5.3-FlashX-Legal:

Billable line (Z.ai)Per 1M tokens
Input$0.37
Cached Input$0.075
Cached Input StorageLimited-time Free
Output$1.25

Input and output match at $0.37 / $1.25. On the listed pages checked 2026-11-03 ~14:36 Asia/Shanghai, OpenRouter Cache read and Z.ai Cached Input both show $0.075 per 1M. Final settlement follows the bill at call time—do not treat an older $0.09 screenshot as current. Z.ai also labels Cached Input Storage as limited-time free; OpenRouter’s model page did not post that storage row as a separate rate. OpenRouter’s FAQ for the same model restates the $0.37 / $1.25 / Cache Read $0.075 listed band.

Context, max completion, and reasoning

Context length is 1,048,576 on the OpenRouter FAQ and the models API. Max completion diverges across vendor pages:

SourceCompletion / max-output figure
OpenRouter FAQ + API top_provider.max_completion_tokens131,072
Z.ai Flash / FlashX overview card (Maximum Output Tokens)128K

We haven't measured either figure in a live run. Both appear in vendor docs; both stay listed until the pages converge or a measured accepted ceiling lands.

Reasoning on this OpenRouter id is mandatory and cannot be disabled. Supported efforts: low, high, max; default max. Z.ai’s Flash / FlashX docs likewise say thinking.type only supports enabled. The models API also lists tools, structured outputs, temperature, top-p, top-k, and related parameters among supported_parameters.

How Z.ai docs place FlashX Legal

Z.ai’s GLM-5.3-Flash / FlashX guide lists Model Codes glm-5.3-flash, glm-5.3-flashx, and glm-5.3-flashx-legal. The GLM-5.3-FlashX-Legal section (also GLM-Legal) frames professional legal workflows: research, contract review, and document drafting, preferably inside an agent environment with specialized MCP services, with AutoClaw Legal Assistant called out as a ready-made workspace. Z.ai cites Harvey LAB (all-criteria pass rate 19.2%, +12.5 points over the base model, described as SOTA among open-source models in that evaluation)—vendor self-report; we haven't retested.

The Z.ai documentation page did not contain a hard sentence that direct Legal API access still requires a special application. Whether a given Z.ai account can call glm-5.3-flashx-legal without extra approval follows the Z.ai website and console at call time.

Quick facts

  1. Listing time. created 2026-11-02 13:29:14 UTC ＝ 21:29:14 Asia/Shanghai; page Released Nov 2, 2026.
  2. Formal id z-ai/glm-5.3-flashx-legal; canonical …-20261102.
  3. In/Out $0.37 / $1.25 per 1M on both OpenRouter and Z.ai.
  4. Cache lines match on listed pages. OpenRouter Cache read $0.075; Z.ai Cached Input $0.075 (bill at call time).
  5. Context 1,048,576 / 1.0M.
  6. Max completion docs diverge: OpenRouter / API 131,072; Z.ai card 128K.
  7. Reasoning mandatory; efforts low / high / max; default max.
  8. Z.ai code glm-5.3-flashx-legal (GLM-Legal); legal scenes are vendor-described.
  9. Direct Z.ai access follows the live website and console—not inferred from absent “apply first” wording.
  10. We haven't tested prompts, billing lines, or legal task quality on this id.

OpenRouter’s Terra Pro listing (linked above) is a convenience id that pins a reasoning mode on the same underlying Terra. FlashX Legal is a separate Z.ai model id with legal-domain training copy, not a reasoning.mode alias of plain FlashX.

In short

z-ai/glm-5.3-flashx-legal is live on OpenRouter as of 2026-11-02 (~21:29 Beijing time), with canonical …-20261102, listed In/Out $0.37 / $1.25, and listed cache read/cached input both $0.075 on OpenRouter and Z.ai (bill at call time). Context is 1M; completion caps appear as 131,072 on OpenRouter and 128K on Z.ai’s overview card. Reasoning cannot be disabled. Legal-workflow copy and Harvey LAB numbers stay vendor claims until independently measured. Listed rates and caps follow the OpenRouter model page, models API row, Z.ai Flash guide, and Z.ai pricing table at the check times above.