Applied artificial intelligence, neural systems, and smart agents.
Fusing generative AI, multi-agent frameworks, and high-throughput data processing to automate business workflows.
GLM-5.2 Max
GLM-5.2 records a score of 1588 on the WebDev Arena leaderboard and is priced at $1.4 in, $4.4 out — under a fifth of what Opus 5 charges for output. Its total score in this reference coding table is 20.1, and the section below explains why that figure reflects measurement-data limitations more than model quality.
- 1588 on WebDev Arena
- $4.40 output token rate
- 1M token window capacity
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
Max is a parameter, not the name of a model
The Arena leaderboard records an entry with the id glm-5.2-max. Z.ai own technical documentation names that model glm-5.2 and lists max as one of the values of the reasoning_effort parameter. So what earned the ranking is a configuration setting applied to the base model, not a standalone entity.
The same point is recorded on the Kimi page, but there the only statement possible was that Moonshot documents no such model in its specification. Here the vendor exposes the parameter directly in its technical documentation, which makes this page the clearest evidence for the claim: any comparison presenting GLM-5.2 Max as a separate product with its own price has mistaken a leaderboard entry for a model catalogue.
A five-times value ratio, against a score of 1.2
Opus 5 charges $25 per million output tokens; GLM-5.2 charges $4.40. Divide the leaderboard rating by the output rate and GLM lands at 360.9 against Opus 5 at 67.7. On that ratio, GLM is more than five times ahead.
Yet in this reference score-per-dollar column, it draws only 1.2 of twenty weight points. The reason is DeepSeek V4 Flash, which records 5660 in the same column and stretches the scale far enough that the gap between GLM and Opus 5 is compressed flat near the bottom. Min-max normalization behaves this way and is not adjusted in this report, but the column should not be read as placing GLM anywhere near Opus 5 on value per dollar. It is not close — the real gap is far larger than this column shows.
Two empty columns are what hold its rank down
GLM-5.2 totals 20.1 out of 100 in this reference coding table, and fifteen of those hundred points carry no evidence at all. One is the general text leaderboard: no GLM entry has been measured there, so those ten points stay empty. The other is Iran reachability: Z.ai publishes no supported-countries list, and no direct measurement from an Iranian connection has been run.
So the 20.1 figure should be read this way: on the largest column in this table, the web-dev leaderboard, GLM recorded a score of 1588, which is 104 points below the top of the table and 34 points above the floor. The rest of the figure comes from missing measurement data, not from actual evaluation. Once those two columns are filled, the rank shifts on its own, with no manual adjustment applied.
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- not checked
- Payment
- not checked
- Free tier
- no
How we checked: Z.ai publishes no supported-countries list. Until a direct measurement is run from an Iranian connection, this status remains recorded as unmeasured.
Documented strengths
- A $4.40 output rate, under a fifth of Opus 5
- A one-million-token window capacity with a 128k output ceiling, level with the pricier models in this table
- The vendor specifies the reasoning_effort parameter transparently, so max-mode behavior is reproducible
- More than five times ahead of Opus 5 on rating per output dollar
Room for improvement and technical considerations
- On the web-dev leaderboard it sits 104 points below the top of the table and 34 above the floor.
- Its entry on the general text leaderboard has not been measured, so ten points of this table stay empty for it.
- Z.ai publishes no supported-countries list, so the Iran column is recorded as unmeasured.
- It scores zero in the tooling maturity column: no official command-line tool and no editor extension in the specification.
- The 20.1 total reflects measurement-data gaps more than model quality, which makes direct comparison against the leaders unreliable.
Technical verdict
For a high-volume workload where token cost is the deciding factor and top-tier quality is not required, GLM-5.2 is a serious option. But until its text-leaderboard entry and Iran reachability are measured, this model is not recommended for a deployment built around a single model: a choice with half its specification unverified is not a reliable choice.
Frequently raised questions, with documented answers
What is the pricing rate for GLM-5.2
One dollar forty per million input tokens and four dollars forty per million output tokens, per Z.ai official pricing page.
What does reasoning_effort set to max mean
This setting instructs the model to reason more before answering the same request. The underlying model does not change, only the volume of processing before the response increases, and the glm-5.2-max entry on the leaderboard represents exactly this mode.
Why does GLM score so low in this reference table
Because fifteen of a hundred points carry no measurement evidence: its general text leaderboard entry and the Iran reachability column. On the web-dev leaderboard column it recorded a score of 1588, putting its real distance from the leader at 104 points.
Is GLM reachable from Iran
Unmeasured. Z.ai publishes no supported-countries list and no direct measurement from an Iranian connection has been run, so any firm answer at this stage would be a guess.