GLM-4.6

Sep 2025
GLM-4.6
Compare in
on
—per task?
$0.60 / $2.20per 1M in / out?cache $0.11
35 tok/sfirst token in 4.5 s?
200Kcontext

Rankingin non-thinking mode?

Strong at

B
Tool calling#85 of 242Knowledge#153 of 318
C
Agents & work tasks#105 of 175Reasoning#258 of 334Science#215 of 312Long context#222 of 273Instructions#172 of 245

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperCommand A+thinkingOpen×4.8-40%$0.30 / $1.50Context: 200K → 128KBetterGranite 4.2 3BthinkingOpen×2.6-95%$0.006Context: 200K → 128KOOpenMiniCPM5 1Bnon-thinking×75——Context: 200K → 131KCheaperMiniMax M3thinkingOpen×3.7-48%$0.51No strict JSON+1MCheaperQuasar 438B×3.2-10%$2.02

Run it yourself

Fits onhardware needs not verified yet
Licencenot verified

History

10 versions
20406080100Oct 25Jan 26Apr 26Jul 26Release dateAA Omniscience · Hallucination Rate (lower is better)4.54.5V4.64.6V4.755.15.25.3
20406080100Jan 26Jul 26Release dateAA Omniscience · Hallucination Rate (lower is better)4.54.5V4.64.6V4.755.15.25.3

Benchmarks

all categories
Chat & Knowledge
Humanity's Last Exam5.5#260
AA Omniscience Index-31.7#154
CritPt (physics)0#179
AA Intelligence Index 4.3.2≈14.9#145
AA Omniscience · Accuracy21.4#140
AA Omniscience · Hallucination Rate67.6#126
AA-LCR (long context) 1.126.3#224
GPQA Diamond63.2#217
AIME 2025 · AA archived44.3#91
Coding
Terminal-Bench Hard (version unstated)28.8#62
LiveCodeBench · AA archived56.1#78
Agents & Tool Use
τ²-bench · AA76.9#87
Writing & Language
IFBench (instructions)36.7#174

FAQ

How much does GLM-4.6 cost?›

$0.60 per 1M input tokens and $2.20 per 1M output tokens; cached input $0.11 per 1M; cheapest via DeepInfra at $0.40 / $1.60.

Is GLM-4.6 free to use?›

There is no free API; the cheapest paid price is $0.40 per 1M input tokens. The weights are open, so you can run it yourself.

How good is GLM-4.6, and what is it best at?›

#258 of 334 in Chat & Knowledge, #62 of 239 in Coding, #105 of 175 in Agents & Tool Use, #172 of 245 in Writing & Language.

Is GLM-4.6 better than KAT-Coder-Pro V1?›

On AA Omniscience · Hallucination Rate, GLM-4.6 scores 67.6 vs 67.4 — about the same.

How can I use GLM-4.6?›

Through the API of DeepInfra, Featherless, Novita, Z.AI.

What is the context window of GLM-4.6?›

200K tokens.

When was GLM-4.6 released? Is there a newer version?›

Released 2025-09-30. It is the current version in its line.

Is GLM-4.6 open source?›

Yes, the weights are open.

Can I run GLM-4.6 locally?›

The weights are open and can be self-hosted; hardware needs are not confirmed yet.

Can GLM-4.6 read images?›

No, it accepts text only.

What is GLM-4.6?›

GLM-4.6 is an open-weights language model from Z.ai, released Sep 2025.

What are the best alternatives to GLM-4.6?›

Command A+ (cheaper), Granite 4.2 3B (better), MiniCPM5 1B (open weights).