Skip to content

Interface language: EN. Editorial content: German. Prices refer to Germany.

Sir CompareComparisons with evidenceAsk Sir CompareAsk
Germany · EUR
Country and pricesUnited KingdomGBPUnited StatesUSDItalyEURGermanyEURSpainEURFranceEUR
Language: English
ITENDEFRES

Product

GLM-4.6: Datenblatt, API-Preise und Daten

Updated

Z.ai GLM-4.6 — drawn card, not a photograph of the product

Z.aiAI Models & LLMs★ Best for Output cost per 1M tokens

74.6/100

#2 of 3 · AI Models & LLMsThe same yardstick for every verified model in the category.

Z.ai GLM-4.6 leads on Output cost per 1M tokens among the models we verified.

Where it leads

  • Input cost per 1M tokens
  • Output cost per 1M tokens

Where it trails

  • SWE-bench Verified (resolved %)
SWE-bench accuracy
68.2%
Context window
200k tokens
Maximum output
128k tokens
Input cost per 1M tokens
$0.60
Output cost per 1M tokens
$2.20
Open weights (downloadable)
Yes

You use it and pay per use through the vendor's API, per million tokens: there is no shop. The list prices are in the data sheet below.

Official page: docs.z.ai ↗Official page: huggingface.co ↗

Data sheet · GLM-4.6

Every figure is quoted from the vendor's official page, with the date we read it. If a row is missing, the source does not state it and we do not estimate it.

Performance

SWE-bench accuracy
68.2%Rank 4 of 5
What it means and the source

How many real programming problems (issues of open source projects) it solves on its own in the independent SWE-bench Verified ranking. The higher, the better it writes and fixes code.

swebench.comOpens the source in a new tab Oct 28, 2026

Memory and answer length

Context window
200k tokensRank 20 of 21
What it means and the source

How much text the model keeps in view in a single request, question and answer included. It matters for long documents and whole code projects. A token is a fragment of a word.

docs.z.aiOpens the source in a new tab Oct 23, 2026
Maximum output
128k tokensRank 3 of 16
What it means and the source

The longest answer it can write in one go: useful for reports, long code or translations of whole documents.

docs.z.aiOpens the source in a new tab Oct 23, 2026

API prices

Input cost per 1M tokens
$0.60Rank 4 of 21
What it means and the source

What it costs to have the model read a million tokens (your question, the documents, the code). It is the official list price, before discounts.

docs.z.aiOpens the source in a new tab Oct 23, 2026
Cached input cost per 1M tokens
$0.11Rank 10 of 20
What it means and the source

The reduced price when part of the request is identical to one sent recently, for example the same instructions or the same document: the vendor keeps it in memory and charges less for it.

docs.z.aiOpens the source in a new tab Oct 23, 2026
Output cost per 1M tokens
$2.20Rank 4 of 21
What it means and the source

What a million tokens written by the model cost. It usually costs more than the input and is the figure to watch when answers are long.

docs.z.aiOpens the source in a new tab Oct 23, 2026

Size and openness

Parameters
357 billion
What it means and the source

The size of the network: how many billions of learned values make it up. Bigger does not necessarily mean better, but it tells how much hardware it takes to run it yourself.

huggingface.coOpens the source in a new tab Oct 23, 2026
Open weights (downloadable)
Yes7 models out of 23
What it means and the source

The model's weights can be downloaded: you can run it on your own servers, adapt it and decide where the data stays, within the licence terms.

huggingface.coOpens the source in a new tab Oct 23, 2026

Where it stands · GLM-4.6

The same verified figures, next to those of the other catalogue models. Tap a model for the head-to-head.

SWE-bench accuracyRank 4 of 5 · higher is better
  1. 1Claude Opus 4.579.2%
  2. 2Claude Opus 4.675.6%
  3. 3GLM-572.8%
  4. 4GLM-4.668.2%
  5. 5Claude Haiku 4.566.6%
Context windowRank 20 of 21 · higher is better
  1. 1GPT-5.6 Sol1.05M tokens
  2. 1GPT-6 Astra1.05M tokens
  3. 1GPT-6 Luna1.05M tokens
  4. 1GPT-6.1 Sol1.05M tokens
  5. 5Gemini 3.8 Flash1,048,576 tokens
  6. 20GLM-4.6200k tokens
Maximum outputRank 3 of 16 · higher is better
  1. 1DeepSeek V4 Pro384k tokens
  2. 1DeepSeek V4.1 Flash384k tokens
  3. 3GLM-4.6128k tokens
  4. 3Claude Fable 5.1128k tokens
  5. 3Claude Haiku 5.5128k tokens
  6. 3Claude Opus 5.5128k tokens
Input cost per 1M tokensRank 4 of 21 · lower is better
  1. 1Claude Haiku 5.5$0.10
  2. 1GPT-6 Luna$0.10
  3. 3DeepSeek V4.1 Flash$0.30
  4. 4GLM-4.6$0.60
  5. 5Gemini 3.8 Flash$0.75
  6. 6Claude Haiku 4.5$1.00
Output cost per 1M tokensRank 4 of 21 · lower is better
  1. 1Claude Haiku 5.5$0.50
  2. 1GPT-6 Luna$0.50
  3. 3DeepSeek V4.1 Flash$1.20
  4. 4GLM-4.6$2.20
  5. 5GLM-5$3.20
  6. 6Gemini 3.8 Flash$3.75

Hardware to try AI at home

GLM-4.6: just holding the weights in memory takes about 714 GB at 16-bit, 357 GB at 8-bit, 178.5 GB at 4-bit. Beyond 24 GB it is not a home-PC job: using it through an API makes more sense.Worked out from the billions of parameters the vendor states, not a measurement: context and the runtime need more memory.

Questions and answers

GLM-4.6: how much does it cost through the API?

The official list price is $0.60 per 1M input tokens and $2.20 per 1M output tokens.

GLM-4.6: how much text can it read at once?

The context window is 200k tokens. It can write answers of up to 128k tokens.

GLM-4.6: can I run it on my own computer?

Yes, the vendor publishes the model's weights. Parameters: 357 billion. Holding them in memory takes about 178.5 GB at 4-bit; context and the runtime need more memory.

GLM-4.6: how good is it at coding?

It solves 68.2% of the problems in SWE-bench Verified, the independent ranking of real programming problems.

Kurz gefasst

Bei Z.ai GLM-4.6 ist SWE-bench-Genauigkeit der Wert, auf den es ankommt: 68.2. Ich habe sie von https://www.swebench.com/verified übernommen. Unter den 5, die ich geprüft habe, sind 3 besser. Die andere Zahl, auf die es ankommt, ist Kontextfenster: 200K. Bei den messbaren Feldern kommt es auf 74.6/100.

Die Zahlen zu Z.ai GLM-4.6

Ich habe sie nach Gewicht geordnet, mit dem Wert und dem Platz unter den anderen, die ich geprüft habe.

SWE-bench-Genauigkeit: 68.2. Platz 4 von 5, die ich verglichen habe.

Kontextfenster: 200K. Platz 20 von 21, die ich verglichen habe.

Kosten pro 1 Mio. Ausgabe-Token: 2.2. Platz 4 von 21, die ich verglichen habe.

Kosten pro 1 Mio. Eingabe-Token: 0.6. Platz 4 von 21, die ich verglichen habe.

Kosten für Cache-Eingabe pro 1 Mio. Token (USD): 0.11. Platz 11 von 20, die ich verglichen habe.

Maximale Antwort (k Token): 128K. Platz 3 von 16, die ich verglichen habe.

Subject illustration: Illustration of an application (illustrative image, not a photo of the product)
Für wen es sich lohnt

Nimm es, wenn Open weights (downloadable) für dich Vorrang hat: 1 — keines der geprüften Modelle ist besser.

Sieh dich anderswo um, wenn dir SWE-bench-Genauigkeit wichtig ist: hier sind es 68.2, und 3 Produkte sind besser.

Verwandte Inhalte

Quellen

Ask Sir Compare:Explain these numbers to me

How I worked out the score

The same yardstick for every product in this category. Below you can see how much each criterion weighs and what it contributed.

Z.ai GLM-4.6: 74.6/100
  • SWE-bench Verified (resolved %): 90.2/100 × 0.55 = 49.6
  • Output cost per 1M tokens (USD): 100/100 × 0.15 = 15
  • Context window (k tokens): 20/100 × 0.25 = 5
  • Input cost per 1M tokens (USD): 100/100 × 0.05 = 5

Reviews and tests

Z.ai GLM-4.6

Reviews and tests · Z.ai GLM-4.6

Reviews are not available yet.

What changed

  • Page updated

How I put products in order

Sir Compare’s letter

Updates on AI Models & LLMs? I’ll send them your way.

Pick what you care about: I only write when something new comes out — a guide, a news story, a verified price that drops.

Step 1 of 2

What are you into?

Pick as many as you like. Pick none and you get what’s new in every section.

What you want to receive
  • Only what you pick
  • Never an empty email
  • One click to leave

Your assistant

Open full page →