GLM-5.2 API coming soon

Z.ai's open flagship — the #1 open-weight model on the Artificial Analysis Intelligence Index, with a 1M-token context. Arriving at Khidi at half the vendor's price. Kimi K2.6 and MiniMax M2.7 are live today.

$0.70 /1M input $2.20 /1M output Z.ai charges $1.40 / $4.40 · Claude Opus 4.8 charges $5 / $25

How close to the frontier? This close.

Exact versions named; sources: official scorecards and public leaderboards, June 2026.

FrontierSWE

Frontier-grade software engineering. The gap to the best closed model: 0.7 points. The price gap: ~6× at the vendor, ~11× at Khidi.

Claude Opus 4.8
75.1
GLM-5.2
74.4
GPT-5.5
72.6

Code Arena: Frontend (Elo, human-judged)

Real web-development tasks judged head-to-head. Only one model in the world scores higher — Claude Fable 5.

Claude Fable 5
1665
GLM-5.2 (Max)
1595
best Claude Opus variant
<1595
Arena.ai leaderboard, 25 Jun 2026 (bar length ∝ Elo). GLM-5.2 took a 55.0% win share vs Opus 4.7 (Thinking) in matched pairings.

SWE-Bench Pro

Ahead of GPT-5.5. (Opus 4.8's published runs aren't directly comparable on this benchmark.)

GLM-5.2
62.1
GPT-5.5
58.6
On the Artificial Analysis Intelligence Index, GLM-5.2 scores 51 — 4th overall behind only Claude Fable 5, Claude Opus 4.8 and GPT-5.5 (xhigh), and 1st among all open-weight models.

Cost when it lands

Provider50M in + 10M out / month
GPT-5.5 ($5 / $30)$550.00
Claude Opus 4.8 ($5 / $25)$500.00
GLM-5.2 at Z.ai ($1.40 / $4.40)$114.00
GLM-5.2 at Khidi ($0.70 / $2.20)$57.00

Be first in line for GLM-5.2

Sign up now — Kimi K2.6 and MiniMax M2.7 work today with a free trial budget, and account holders get notified the moment GLM-5.2 opens.

Get your API key →