GLM — MIT-licensed weights that beat closed models on security.
Z.ai (formerly Zhipu AI) builds the GLM family under MIT licenses. GLM-5.3 beat Claude Mythos 5 on the CyberGym security benchmark (84.5 vs 83.8), and GLM-5.3 Flash is the cheapest serious multimodal model on the market at $0.15/1M with native vision and video.
Rankings refresh dailyScored on 6 criteriaNo paid rankings
GLM-5.3 beats Claude Mythos 5 on CyberGym security (84.5 vs 83.8)
GLM-5.3 Flash: native vision + video, MIT weights, $0.15/1M input
GLM Coding Plan from $18/month — the cheapest serious coding subscription
3 models
All Z.ai Models
Every Z.ai model in the directory, ranked by overall capability score.
Z.aiBudget
GLM-5.3 Flash
A 320B-A18B mixture-of-experts model with native vision and video, MIT-licensed weights and a 1M context — at fifteen cents per million input tokens.
Verdict
Native vision and video, MIT weights, fifteen cents per million.
Quality score
81%
Pricing
$0.15/1M in
$0.50/1M out
Speed
Fast
4/5 speed
Context
1M tokens
Released August 26, 2026. 320B total parameters, 18B active (320B-A18B MoE). Z.ai lists $0.15/1M input, $0.03/1M cached input and $0.50/1M output, with a launch promotion halving those rates through September 9, 2026. Self-reported against GLM-5.2: DeepSWE 63.4 vs 46.2, AutomationBench 48.8 vs 26.2.
Open weightsMultimodalBudget1M context
Best for
Cheap multimodal work at scale on MIT-licensed weights
Z.ai's newest flagship, aimed squarely at software engineering, autonomous agents and cybersecurity — and the first open-weights model to beat Claude Mythos 5 on a security benchmark.
Verdict
Same price as GLM-5.2, far stronger on agents and security.
Quality score
81%
Pricing
$1.40/1M in
$4.40/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Released August 14, 2026. Z.ai list pricing is $1.40/$4.40, the same rate as GLM-5.2; resellers discount from that list. Also available through the GLM Coding Plan from $18/mo. Reported GPQA Diamond 91.7% and Artificial Analysis Intelligence Index 59.5.
Open weightsAgenticSecurity1M context
Best for
Agentic engineering and security work on open weights
Z.ai's MIT-licensed open-weight flagship — the top open-weights coding model of mid-2026, beating GPT-5.5 on agentic coding benchmarks at roughly a sixth of the cost.
Verdict
Top open-weights coder — beats GPT-5.5 at a sixth of the cost.
Quality score
80%
Pricing
$1.40/1M in
$4.40/1M out
Speed
Balanced
3/5 speed
Context
1M tokens
Announced June 13, 2026; pay-per-token API live June 16. Two reasoning modes ('thinking' and 'max thinking'). GLM Coding Plan: Lite $18/mo (~$12.60 effective yearly), Pro $72, Max $160.
Pricing changes, new releases, and ranking shifts — straight to your inbox.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
Z.ai FAQ
What is Z.ai's best AI model in 2026?
GLM-5.3 ($1.40/$4.40) — huge agentic gains over GLM-5.2 at the same price (DeepSWE 46.2 → 66.9) and a first for open weights: it beats Claude Mythos 5 on the CyberGym security benchmark. GLM-5.3 Flash ($0.15/$0.50) is the multimodal value pick with native vision and video.
Are GLM models open source?
Yes — GLM-5.2, GLM-5.3, and GLM-5.3 Flash all ship MIT-licensed weights, the most permissive license among frontier-adjacent labs. You can self-host, fine-tune, and deploy commercially with no restrictions, or use the first-party API and GLM Coding Plan (from $18/month).
How does GLM compare to Claude and GPT?
On security benchmarks GLM-5.3 actually leads — 84.5% on CyberGym vs Claude Mythos 5's 83.8%. On general agentic coding it sits below the closed frontier (no published SWE-bench Verified score) but well above its price class: Terminal-Bench 2.1 88.2 rivals Grok 4.6 at a third of the cost.
What is the GLM Coding Plan?
Z.ai's subscription for coding-tool usage: Lite at $18/month (~$12.60 effective on yearly billing), Pro at $72, and Max at $160. It routes GLM models into Claude Code-compatible tooling and is the cheapest serious coding subscription in 2026.