Models / Z.AI/ GLM 5.2

GLM 5.2

Name: GLM 5.2
Author: Z.AI

Z.AI · released Jun 16, 2026 · zai-org/GLM-5.2

InputOutput

Type: Open weightsMIT License
Params: 753B
Context: 1M

about 786K words of context

Our take

Written Aug 2, 2026

Available from more hosts than any other model we track, this large downloadable model from Zhipu carries a permissive MIT licence and a one-million-token request limit. It is a strong choice for teams that want top-tier capability without licensing restrictions.

Who should pick it

Pick this when you need top-tier capability with a genuinely permissive licence, or for long-context workloads on a budget. Use it if you want leverage over providers: 30 offers means real price competition. Skip it if you need image, audio or video input, or if you want to self-host on a single consumer GPU.

The case for it

Permissive MIT licence allows commercial use, fine-tuning and redistribution.
30 current hosted offers, so price competition is unusually strong.
One-million-token request limit matches the proprietary top-tier models.

The case against it

Text-only; no image, audio or video input, unlike Gemini or Claude lines.
Over 753.3 billion parameters puts even compressed weights beyond workstation reach.

How good is it?

What has been measured, on which board, and where GLM 5.2 placed.

IntelligencePuzzles, maths, exam questions

4 of 5

Arena Text (overall)18th of 143 · 1469.5via Max

Arena Hard Prompts 21st of 143 via MaxArena Maths 19th of 139 via MaxLiveBench Mathematics 15th of 35LiveBench Data Analysis 16th of 35LiveBench Reasoning 25th of 35

CodingWriting and fixing code on its own

4 of 5

Arena Coding29th of 143 · 1505.7via Max

Arena Code (WebDev) 5th of 74 via MaxLiveBench Coding 11th of 35

AgenticPlanning, calling tools, staying on task

3.5 of 5

Arena Agent (IPS)7th of 36 · 0.071via Max

LiveBench Agentic Coding 11th of 35

WritingWe do not rate this

Scored, not ratedThe placings are on the right.

Two boards come close and neither tests writing: Arena Creative Writing asks people which of two replies they prefer, and LiveBench Language tests whether a model understood a passage. So we show where GLM 5.2 placed and give it no mark out of five.

Arena Creative Writing 23rd of 143 · 1444.6 via MaxLiveBench Language 20th of 35 · 76.2

Also scored, on boards we give no mark for

Arena Instruction Following 18th of 143 via MaxLiveBench 19th of 35LiveBench Instruction Following 28th of 35

These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done, which is why they get no rating.

Every published score for this model16 scoresEvery figure we hold, from 16 boards, with who ran it and a link to the source — including the boards no rating above is built on.

LiveBenchreasoning

73.2independentsource ↗

LiveBench Agentic Codingagentic

51.8independentsource ↗

LiveBench Codingcoding

79.7independentsource ↗

LiveBench Data Analysisreasoning

73.7independentsource ↗

LiveBench Instruction Followingchat

62.3independentsource ↗

LiveBench Languagechat

76.2independentsource ↗

LiveBench Mathematicsmath

89.8independentsource ↗

LiveBench Reasoningreasoning

78.6independentsource ↗

Arena Agent (IPS)agentic

0.071via Maxindependentsource ↗

Arena Codingcoding

1505.7via Maxindependentsource ↗

Arena Creative Writingchat

1444.6via Maxindependentsource ↗

Arena Hard Promptsreasoning

1488via Maxindependentsource ↗

Arena Instruction Followingchat

1464via Maxindependentsource ↗

Arena Mathsmath

1475.1via Maxindependentsource ↗

Arena Text (overall)chat

1469.5via Maxindependentsource ↗

Arena Code (WebDev)coding

1586.3via Maxindependentsource ↗

Can you run it yourself?

Fits fully in memory on 0 of the 71 devices we track, at Q4.

Fits in memory: weights load entirely on the card
Spills to system RAM: some weights offload; much slower
Too large: will not load even with offload
est: size is calculated; the verdict could change by 10%

A card many people ownToo large

GeForce RTX 4090 · 24 GB

Weights at Q4_K_M475 / 24 GBest

Usable contextNot calculated for spilled setups

Decode speedNot estimated for spilled setups

Too large for this card. The weights do not fit even with part of them offloaded to system memory.

One step upToo large

Apple M1 Pro (16-core GPU) · 32 GB

Weights at Q4_K_M475 / 32 GBest

Usable contextNot calculated for spilled setups

Decode speedNot estimated for spilled setups

Too large for this card. The weights do not fit even with part of them offloaded to system memory.

One step downToo large

Radeon RX 7900 XT · 20 GB

Weights at Q4_K_M475 / 20 GBest

Usable contextNot calculated for spilled setups

Decode speedNot estimated for spilled setups

Too large for this card. The weights do not fit even with part of them offloaded to system memory.

Your hardware

Checking your profile…

Memory use by level

Against a 24 GB card.

Q4_K_M

recommended

475 GBest

Too large

Q5_K_M

557.2 GBest

Too large

Q8_0

832.4 GBest

Too large

All 3 sizes here are calculated, not measured. We hold no measured file for this model, so each size comes from the parameter count and every verdict above inherits that uncertainty.

Check against your own machine → · All 71 devices, with every size →

Or rent it from someone else

31 providers, from 40 listings.

Cheapest published offer

OpenRouter

Cheapest of 40 live listings. Picked at the widest standard context we hold, within one quantisation slice, so the numbers beside it are a price one host actually charges.

per 1M tokens: $0.76 in / $2.42 out
Context served: 1M
Throughput: Not measured

Current provider offers with price, context and prompt-privacy answers
Provider	In / out per 1M tokens	Context	Throughput	Trains on prompts	Logs prompts	Zero retention
Decartfp4	$0.60 / $1.50	1M	37 tok/s	No	No	Confirmed
Novita AIfp8	$0.63 / $1.98	1M	23 tok/s	No	No	Confirmed
StreamLakefp8	$0.63 / $1.98	1M	31 tok/s	No	Yesunknown period	Unknown
Baidufp8	$0.76 / $2.38	1M	46 tok/s	No	Yesunknown period	Unknown
DeepInfrafp4	$0.75 / $2.40	1M	not measured	Unknown	Unknown	Unknown
DeepInfrafp4	$0.75 / $2.40	1M	35 tok/s	No	No	Confirmed
CoreWeavefp4	$0.76 / $2.42	262K	100 tok/s	No	No	Confirmed
OpenRouter	$0.76 / $2.42	1M	not measured	Unknown	Unknown	Unknown
Ambientfp8	$0.76 / $2.42	101K	not measured	No	Yesunknown period	Unknown
AkashMLfp8	$0.77 / $2.42	97K	36 tok/s	No	No	Confirmed
Decartfp8	$1.20 / $2.50	1M	not measured	No	No	Unknown
Alibaba Cloud	$0.83 / $2.60	1M	42 tok/s	No	Yesunknown period	Unknown
Inceptronfp4	$0.94 / $2.90	1M	17 tok/s	No	No	Confirmed
GMICloudfp8	$0.92 / $2.90	1M	35 tok/s	No	Yesunknown period	Unknown
Alibaba Cloudfp8	$0.97 / $3.04	1M	44 tok/s	No	Yesunknown period	Unknown
Sail Researchfp8	$1.00 / $3.50	1M	54 tok/s	No	No	Confirmed
SiliconFlowfp8	$1.19 / $3.74	1M	33 tok/s	No	No	Confirmed
Chutesfp4	$1.25 / $3.95	1M	26 tok/s	No	Yesunknown period	Unknown
AtlasCloudfp8	$1.26 / $3.96	1M	26 tok/s	No	Yesunknown period	Unknown
Phala	$1.26 / $3.96	1M	28 tok/s	No	No	Confirmed
Waferfp4	$1.26 / $3.96	1M	74 tok/s	No	No	Unknown
Morph	$1.10 / $4.10	1M	46 tok/s	No	No	Confirmed
Ionstreamfp4	$1.40 / $4.40	1M	77 tok/s	No	No	Confirmed
Novita AI	$1.40 / $4.40	1M	not measured	Unknown	Unknown	Unknown
Crusoefp8	$1.40 / $4.40	1M	113 tok/s	No	No	Confirmed
DigitalOcean Gradient	$1.05 / $4.40	262K	48 tok/s	No	No	Confirmed
Parasailfp4	$1.40 / $4.40	262K	112 tok/s	No	No	Confirmed
Basetenfp8	$1.40 / $4.40	1M	65 tok/s	No	No	Confirmed
Cloudflare Workers AI	$1.40 / $4.40	262K	78 tok/s	No	Yesunknown period	Unknown
Fireworks AI	$1.40 / $4.40	1M	45 tok/s	No	No	Confirmed
Friendli	$1.40 / $4.40	1M	94 tok/s	No	Yesunknown period	Unknown
Z.AIfp8	$1.40 / $4.40	1M	17 tok/s	No	No	Confirmed
Venice AIfp8	$1.40 / $4.40	1M	43 tok/s	No	No	Confirmed
Together AI	$1.40 / $4.40	512K	85 tok/s	No	No	Confirmed
Waferfast tierfp4	$2.10 / $6.60	1M	95 tok/s	No	No	Unknown
Cloudflare Workers AIfast tier	$2.10 / $6.60	262K	66 tok/s	No	Yesunknown period	Unknown
Fireworks AIfast tier	$2.10 / $6.60	1M	93 tok/s	No	No	Confirmed
Basetenfast tierfp8	$2.10 / $6.60	524K	135 tok/s	No	No	Unknown
Alibaba Cloudfast tierfp8	$2.31 / $7.26	1M	69 tok/s	No	Yesunknown period	Unknown
Io Netfp8	$3.66 / $8.01	262K	88 tok/s	No	No	Unknown

Across the 40 listings we hold: 37 say they do not train on prompts, 0 say they do and 3 do not say. 20 appear in the zero-retention registry we check; the rest are unknown to us rather than confirmed either way.

What each host's API supports

From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.

✓
Supported: ✗
Not supported
Not published: host gave no parameter list

API features per host
Provider	Tool calling	JSON output	Strict schema
Decartfp4	✓	✓	✓
Novita AIfp8	✓	✓	✗
StreamLakefp8	✓	✓	✓
Baidufp8	✓	✓	✓
DeepInfrafp4
DeepInfrafp4	✓	✓	✓
CoreWeavefp4	✓	✓	✓
OpenRouter	✓	✓	✓
Ambientfp8	✓	✓	✓
AkashMLfp8	✓	✓	✓
Decartfp8
Alibaba Cloud	✓	✓	✓
Inceptronfp4	✓	✓	✓
GMICloudfp8	✓	✓	✓
Alibaba Cloudfp8	✓	✓	✓
Sail Researchfp8	✓	✓	✗
SiliconFlowfp8	✓	✗	✗
Chutesfp4	✓	✓	✓
AtlasCloudfp8	✓	✓	✗
Phala	✓	✓	✓
Waferfp4	✓	✓	✓
Morph	✓	✓	✓
Ionstreamfp4	✓	✓	✓
Novita AI
Crusoefp8	✓	✓	✓
DigitalOcean Gradient	✓	✓	✗
Parasailfp4	✓	✓	✓
Basetenfp8	✓	✓	✓
Cloudflare Workers AI	✓	✓	✓
Fireworks AI	✓	✓	✓
Friendli	✓	✓	✓
Z.AIfp8	✓	✓	✗
Venice AIfp8	✓	✓	✓
Together AI	✓	✓	✓
Waferfast · fp4	✓	✓	✓
Cloudflare Workers AIfast	✓	✓	✓
Fireworks AIfast	✓	✓	✓
Basetenfast · fp8	✓	✓	✓
Alibaba Cloudfast · fp8	✓	✓	✓
Io Netfp8	✓	✓	✓

Tool calling: 37 of 40 listings say yes, 3 publish no parameter list. JSON output: 36 of 40 listings say yes, 1 says no, 3 publish no parameter list. Strict schema: 31 of 40 listings say yes, 6 say no, 3 publish no parameter list.

Models people weigh against GLM 5.2

Shortlist any of these to compare them side by side.

Weighed against itKimi K2.7 Code Kimi K2.6 DeepSeek V3.2

Other sizes and versionsGLM 5.1 GLM 5 GLM 4.7 GLM 4.6V

SucceedsGLM 5.1

Compare them field by field →

When we formed this view

And whether anything has moved since.

Dates behind this page

Aug 4, 2026Price changeOpenRouter raised GLM 5.2 pricing by 22%input +21% ($0.63 → $0.76 per 1M tokens); output +22% ($1.98 → $2.42 per 1M tokens); cache read +20% ($0.12 → $0.14 per 1M tokens)detected

Aug 3, 2026Price changeGLM 5.2 cut across 6 hosts, by up to 10% at PhalaGLM 5.2 moved on 6 hosts: Phala: input −10% ($1.40 → $1.26 per 1M tokens); output −10% ($4.40 → $3.96 per 1M tokens); cache read −10% ($0.26 → $0.23 per 1M tokens); StreamLake: input −11% ($0.71 → $0.63 per 1M tokens); output −11% ($2.23 → $1.98 per 1M tokens); cache read −11% ($0.13 → $0.12 per 1M tokens); OpenRouter: input −11% ($0.71 → $0.63 per 1M tokens); output −11% ($2.22 → $1.98 per 1M tokens); cache read −11% ($0.13 → $0.12 per 1M tokens); Novita: input −11% ($0.71 → $0.63 per 1M tokens); output −11% ($2.22 → $1.98 per 1M tokens); cache read −11% ($0.13 → $0.12 per 1M tokens); Decart: input −17% ($0.72 → $0.60 per 1M tokens); output −17% ($1.80 → $1.50 per 1M tokens); cache read −17% ($0.12 → $0.10 per 1M tokens); Chutes: cache read −80% ($0.63 → $0.13 per 1M tokens)detected

Aug 2, 2026BenchmarkScored 1505.7 via Max on Arena Codingleaderboardannounced

Aug 2, 2026BenchmarkScored 1444.6 via Max on Arena Creative Writingleaderboardannounced

Aug 2, 2026BenchmarkScored 1488 via Max on Arena Hard Promptsleaderboardannounced

Aug 2, 2026BenchmarkScored 1464 via Max on Arena Instruction Followingleaderboardannounced

Aug 2, 2026BenchmarkScored 1475.1 via Max on Arena Mathsleaderboardannounced

Aug 2, 2026BenchmarkScored 1469.5 via Max on Arena Text (overall)leaderboardannounced

Aug 2, 2026BenchmarkScored 1586.3 via Max on Arena Code (WebDev)leaderboardannounced

Aug 1, 2026Price changeGLM 5.2 repriced across 5 hosts, from a 24% rise at StreamLake to a 40% cut at DecartGLM 5.2 moved on 5 hosts: StreamLake: input +24% ($0.61 → $0.76 per 1M tokens); output +24% ($1.93 → $2.39 per 1M tokens); cache read +24% ($0.11 → $0.14 per 1M tokens); Baidu: input +8% ($0.70 → $0.76 per 1M tokens); output +8% ($2.20 → $2.38 per 1M tokens); cache read +8% ($0.13 → $0.14 per 1M tokens); Novita: input −6% ($0.76 → $0.72 per 1M tokens); output −6% ($2.38 → $2.25 per 1M tokens); cache read −6% ($0.14 → $0.13 per 1M tokens); OpenRouter: input −6% ($0.76 → $0.72 per 1M tokens); output −6% ($2.39 → $2.25 per 1M tokens); cache read −6% ($0.14 → $0.13 per 1M tokens); Decart: input −40% ($1.20 → $0.72 per 1M tokens); output −28% ($2.50 → $1.80 per 1M tokens); cache read −40% ($0.20 → $0.12 per 1M tokens)detected

Prices last checked 38h ago

What we do not know about this model yet

We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
3 of 40 listings publish no parameter list, so what their API accepts is unknown to us.
Nothing we hold says whether an endpoint streams, so we do not show it either way.
3 of 40 listings do not say whether they train on prompts.

Licence and identifiers

What the licence allowsMIT License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.

Licence

MIT License

permissiveCommercial use allowed

Fully permissive: do anything with attribution. No patent grant, unlike Apache-2.0.

Identifiers

Hugging Face: zai-org/GLM-5.2
Architecture: Mixture of experts
Modality record: text->text
Catalogue slug: z-ai-glm-5-2

Machine-readable model card (omc.json) →

Explore furtherWhere to run GLM 5.2 Best open-weight models Longest context windows

Something wrong on this page? Tell us