Text and chat · Z.ai (Zhipu AI)

GLM

Sub Rosa's default chat model: an open model from Z.ai with strong reasoning and tool use, in private mode.

Private, some versions anonymized Moderate cost
Latest versionGLM 5.3 FlashAugust 26, 2026
Artificial Analysis Intelligence Index45 points#13 of 90 on Artificial Analysis Intelligence Index
Price$1.978 per million tokens written$0.00138 a page written
Key spec1M tokens of context

What sets it apart

GLM is an open-weights family with a steady release cadence: six versions in under a year, from 357B to 753B parameters, and a 320B Flash tier that keeps a 1M context window at a fraction of the price. Most versions ship under the permissive MIT license.

  • Open weights, MIT on most versions
  • 753B MoE with 40B active parameters
  • Flash tier: 1M context, reads images

Recommended for: write, summarize and answer questions every day ; keep a sensitive conversation as private as possible.

Strengths and limits

Where it shines

  • Default in Sub Rosa: strong reasoning, tool use and structured output
  • Private mode on every version: zero data retention
  • 1M token context on GLM 5.2 and 5.3, enough for long documents
  • Mid-range price for its level; Flash versions cost a fraction of that

Think twice when

  • GLM 5.2, 5.1 and 5.3 read text only: attach an image and pick another model
  • GLM 5 Turbo and 5V Turbo are anonymized, not private
  • GLM 5.2: 1M token context, text input
  • Tool calling, reasoning effort, JSON schema output
  • Images: GLM 5.3 Flash and GLM 5V Turbo
  • Open weights (MIT license) for GLM 5.2

Use cases

Reading a long contract

Here is a 90-page supplier contract. List every clause that lets the supplier raise prices or end the agreement early, with the page number and a one-line plain-English summary.

GLM 5.2 and 5.3 take a 1M token context, so a long document fits in one message.

Cheap batch work with images

For each screenshot I attach, say which app it shows and write one sentence describing the error message, if any.

GLM 5.3 Flash reads images and costs $0.15 per million input tokens.

Fixing a bug across several files

The checkout total is wrong when a discount code and a gift card are both used. Find the cause in the code I pasted and propose a minimal fix.

GLM 5.3 was tuned after pre-training for coding work, per its model card.

Against its rivals

  • KimiKimi K3 scores about the same (44 vs 45) but costs roughly twice as much per token.#15 of 90 on Artificial Analysis Intelligence Index (44 points)
  • DeepSeekDeepSeek V4.1 Flash is cheaper still; GLM 5.3 scores higher on the Artificial Analysis index.#24 of 90 on Artificial Analysis Intelligence Index (39 points)
  • Qwen Max, Plus and FlashQwen 3.8 Max matches GLM 5.3 on the index (45) but its weights are closed.#14 of 90 on Artificial Analysis Intelligence Index (45 points)
Compare side by side ↗

Version history

Each version, dated, with what it changed over the one before.

Artificial Analysis Intelligence Index, version after versionHigher is better · Artificial Analysis

Hollow points are earlier versions no longer offered in Sub Rosa.

See the numbers
VersionReleasedpoints
GLM 4.62025-09-3015
GLM 4.72025-12-2222
GLM 4.7 Flash2026-01-1915
GLM 52026-02-1128
GLM 5 Turbo2026-03-1527
GLM 5V Turbo2026-04-0123
GLM 5.12026-04-0726
GLM 5.22026-06-1634
GLM 5.32026-08-1845
GLM 5.3 Flash2026-08-2642
  1. GLM 4.6
    • 357B-parameter mixture of experts (32B active) with a 200K context window, under the MIT license.
    Source : artificialanalysis.ai
  2. GLM 4.7
    • Same 357B MoE size, now listed as a reasoning model with extended thinking.
    Source : artificialanalysis.ai
  3. GLM 4.7 Flash
    • Small tier: 31.2B total and 3B active parameters, MIT weights, priced at $0.07 per million input tokens.
    Source : artificialanalysis.ai
  4. GLM 5
    • Jump to 744B parameters (40B active), still with a 200K context and the MIT license.
    Source : artificialanalysis.ai
  5. GLM 5 Turbo
    • Proprietary, hosted-only variant with undisclosed size and a 200K context.
    Source : artificialanalysis.ai
  6. GLM 5V Turbo
    • Adds image and video input to the Turbo line, with text output.
    Source : artificialanalysis.ai
  7. GLM 5.1
    • Refresh of the 744B MoE (40B active), 200K context, MIT license, replacing GLM 5.
    Source : artificialanalysis.ai
  8. GLM 5.2
    • 753B parameters (40B active) and a 1M token context window, five times the previous 200K.
    Source : artificialanalysis.ai
  9. GLM 5.3
    • Same base model as GLM 5.2, with gains from post-training only: about 50% better on Z.ai’s own Code Bench.
    • Moves from MIT to a custom GLM 5.3 license with commercial restrictions.
    Source : huggingface.co
  10. GLM 5.3 FlashLatest
    • 320B MoE with 18B active parameters, 1M context, accepts images, MIT license.
    Source : artificialanalysis.ai

Benchmarks

How to read these scores: what a benchmark tells you.

Artificial Analysis Intelligence Index: #13 of 90 models measured in the catalog. One score averaging about ten independent tests of reasoning, knowledge, math, coding and agent work, all run the same way for every model. Higher is better. Each model is listed at its highest reasoning effort. Two or three points apart is a close call; ten points is a different class of model.

Artificial Analysis Intelligence IndexHigher is better · Artificial Analysis, October 4, 2026
  1. 1. Claude Opus 5.5Claude58
  2. 2. Claude Sonnet 5.5Claude56
  3. 3. Claude Fable 5.1Claude53
  4. 4. GPT-6 AstraGPT53
  5. 5. Gemini 4 ArgonGemini53
  6. 6. GPT-6.1 SolGPT52
  7. 7. Muse WonderMuse52
  8. 8. Claude Opus 5Claude51
  9. 9. Claude Fable 5Claude50
  10. 10. GPT-6 SolGPT48
  11. 13. GLM 5.3GLM45
  12. 18. GLM 5.3 FlashGLM42
  13. 35. GLM 5.2GLM34
See the numbers
Modelpoints
1. Claude Opus 5.558
2. Claude Sonnet 5.556
3. Claude Fable 5.153
4. GPT-6 Astra53
5. Gemini 4 Argon53
6. GPT-6.1 Sol52
7. Muse Wonder52
8. Claude Opus 551
9. Claude Fable 550
10. GPT-6 Sol48
13. GLM 5.345
18. GLM 5.3 Flash42
35. GLM 5.234

All 13 models in this family

ModelUsePrivacyPrice
GLM 5.2zai-org-glm-5-2TextPrivate$1.978 per million tokens written$0.00138 a page written
GLM 5.1zai-org-glm-5-1TextPrivate$2.176 per million tokens written$0.00152 a page written
GLM 5zai-org-glm-5TextPrivate$1.439 per million tokens written$0.00101 a page written
GLM 4.6zai-org-glm-4.6TextPrivate$0.787 per million tokens written$0.000551 a page written
GLM 4.7zai-org-glm-4.7TextPrivate$1.192 per million tokens written$0.000834 a page written
GLM 4.7 Flashzai-org-glm-4.7-flashTextPrivate$0.18 per million tokens written$0.000126 a page written
GLM 5.3z-ai-glm-5-3TextPrivate$2.473 per million tokens written$0.00173 a page written
GLM 5.3 Flashz-ai-glm-5-3-flashTextPrivate$0.225 per million tokens written$0.000157 a page written
GLM 5 Turboz-ai-glm-5-turboTextAnonymized$1.798 per million tokens written$0.00126 a page written
GLM 5V Turboz-ai-glm-5v-turboTextAnonymized$2.248 per million tokens written$0.00157 a page written
GLM 5.2 E2EEe2ee-glm-5-2-pTextPrivate$2.585 per million tokens written$0.00181 a page written
GLM 5.3 E2EEe2ee-glm-5-3-pTextPrivate$2.473 per million tokens written$0.00173 a page written
GLM 5.3 Flash E2EEe2ee-glm-5-3-flashTextPrivate$0.243 per million tokens written$0.00017 a page written

Specifications

SpecificationGLM 5.3 FlashGLM 5.3GLM 5.2GLM 5.1GLM 5V Turbo
ReleasedAugust 26, 2026August 18, 2026June 16, 2026April 7, 2026April 1, 2026
Parameters320B total, 18B active753B total, 40B active753B total, 40B active744B total, 40B activeNot disclosed
WeightsOpen, MITOpen, GLM 5.3 License (commercial restrictions)Open, MITOpen, MITClosed
Acceptstext, imagestexttexttexttext, images, video
Producestexttexttexttexttext
Context1,049K tokens1,000K tokens1,000K tokens200K tokens200K tokens
AbilitiesReads images, Uses Sub Rosa’s tools, Thinks before answering, Tuned for codeUses Sub Rosa’s tools, Thinks before answering, Tuned for codeUses Sub Rosa’s tools, Thinks before answering, Tuned for codeUses Sub Rosa’s tools, Thinks before answeringReads images, Uses Sub Rosa’s tools, Thinks before answering, Tuned for code
Price$0.225 per million tokens written$2.473 per million tokens written$1.978 per million tokens written$2.176 per million tokens written$2.248 per million tokens written
PrivacyPrivatePrivatePrivatePrivateAnonymized

Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.