← the radar · ✳ model tracker

GLM-5.3

● live
Zhipu AIopen weightsChinese labs stream

Live - same 743B base as 5.2, every gain from post-training · tops CyberGym · open weights on Hugging Face (zai-org/GLM-5.3, 141 safetensors files, public when we read it Sep 25; repo created Aug 25) · Anthropic research (Sep 29): builds end-to-end exploits near Claude Mythos Preview's rate (ExploitBench 50 of 410 vs 56 of 410), and its safeguards fell to simple bypasses 64-100% of the time. On the board as live since 2026-08-14 — the watchers keep tracking it: lineup changes, pricing moves and deprecations all alert.

Answering in 1.9s· checked Live model status →

Watch for Zhipu AI’s next model, free →Launch report (PDF)

Timeline

2026-09-29
Anthropic: open-weights GLM-5.3 nearly matches Claude Mythos Preview at building exploits, and its safeguards are easy to strippost

Anthropic published research on September 29 testing Z.ai's GLM-5.3, an open-weights model anyone can download, on offensive cyber tasks. Its finding: GLM-5.3 can build working end-to-end exploits at close to the rate of Claude Mythos Preview, the model Anthropic released only to vetted defenders.

2026-08-14
GLM-5.3 verified live — from the vendor's own announcement and catalog listingverified
2026-08-14
GLM-5.3 tops CyberGym and trails the frontier on exploitation — and we missed it because the sweep had no eyes on models.devpost

Z.ai shipped GLM-5.3 today, and we did not catch it. The entry was armed and the pattern was right — the sweep simply had no source that carried the id. That is now fixed, and the fix is the more useful half of this post.

On the bench index

LMArena: 1,623LiveBench: 76.1SimpleBench: 66.2LLM Stats: 52Vals AI: 53.5Artificial Analysis: 44.8AW Index: 47.8

full comparison table →

Coding leaderboard

#17 of 31 on the coding index, at 46.7 · combines LMArena · Coding, LiveBench · Coding, Vals Vibe Code Bench

the AI coding leaderboard →

Open weights

#2 of 8 open-weights models on the Bench Index · zai-org/GLM-5.3 on Hugging Face · 753B parameters · license custom (glm-5.3)

the open-weights leaderboard →

Where to use GLM-5.3

71 providers list GLM-5.3 in the models.dev catalog (as of 2026-10-04), under the model ids shown below.

Z.AIglm-5.3
Zhipu AIglm-5.3
OpenRouterz-ai/glm-5.3 · z-ai/glm-5.3-prime
Vercel AI Gatewayzai/glm-5.3 · zai/glm-5.3-fast
302.AIglm-5.3
AgentRouterglm-5.3
ai&zai-org/glm-5.3
AIHubMixglm-5.3
AKI.IOglm5.3-754b
Alibaba (China)glm-5.3
Alibaba Token Planglm-5.3
Alibaba Token Plan (China)glm-5.3

Also listed by Baseten, Bothub, Charm Hyper, ClinePass, Cloudflare Workers AI, CoralBricks, Cortecs, CrofAI, CrossModel, Deep Infra, DevPass (LLM Gateway), DigitalOcean, Eden AI, EmpirioLabs AI, engy, Fireworks AI, Friendli, GreenPT, Hugging Face, Inco, IteraCompute, Kenari, Kilo Gateway, LLM Gateway, Melious, Merge Gateway, Mistral, NaN, NanoGPT, Nebius Token Factory, Neuralwatt, Nvidia, Ofox, Ollama Cloud, OpenCode Go, OpenCode Zen, Opper, OrcaRouter, Privatemode AI, Requesty, SCNet Token Plan, SiliconFlow, Synthetic, Tempr Gateway, TensorX, Tinfoil, Together AI, TokenGo, TokenRouter, Umans AI, Umans AI Coding Plan, Vancine, Venice AI, Vivgrid, Volcengine Ark Coding Plan, Vultr, Z.AI Coding Plan, ZenMux, Zhipu AI Coding Plan.

Listings cover the model family, so a provider may list a sibling tier under the same name. Check the id before you build on it.

GLM-5.3: quick answers

Is GLM-5.3 out?

Yes. GLM-5.3 is live and has been since August 14, 2026.

When was GLM-5.3 released?

August 14, 2026.

Is GLM-5.3 good for coding?

On the ArtificialWatch coding index it scores 46.7, #17 of 31 ranked models. The index combines LMArena · Coding, LiveBench · Coding and Vals Vibe Code Bench, taking the median of each model's percentile across them.

Is GLM-5.3 open source?

It has open weights: the trained model is public on Hugging Face as zai-org/GLM-5.3, 753B parameters as Hugging Face counts them, license custom (glm-5.3). Among the 8 open-weights models the ArtificialWatch Bench Index ranks, it is #2.

Who makes GLM-5.3?

Zhipu AI. ArtificialWatch also tracks GLM-5.3 Flash, GLM-5.2, GLM-5.5 and GLM-6 from Zhipu AI.

Where can I use GLM-5.3?

71 providers list it in the models.dev catalog, including Z.AI, Zhipu AI, OpenRouter, Vercel AI Gateway and 302.AI. The exact model ids each one lists are on this page.

How much does GLM-5.3 cost?

Z.AI's own API lists GLM-5.3 (glm-5.3) at $1.40 per million input tokens and $4.40 per million output tokens, and $0.26 per million cached input tokens. Prices as published in the models.dev catalog, as of 2026-10-04.

What is GLM-5.3's context window?

1M tokens, with up to 131K output tokens per response, as Z.AI lists GLM-5.3 (glm-5.3).

Is GLM-5.3 open weights?

Yes. GLM-5.3 is released with open weights.

Every live model's API price, side by side →

Head to head

GLM-5.3 vs Kimi K3GLM-5.3 vs GLM-5.2GLM-5.3 vs GPT-6 AstraGLM-5.3 vs Muse Spark 1.3GLM-5.3 vs Qwen 3.8 MaxGLM-5.3 vs GLM-5.3 FlashGLM-5.3 vs DeepSeek V4GLM-5.3 vs Claude Opus 5GLM-5.3 vs Fable 5

More from Zhipu AI

GLM-5.3 Flashlive since 2026-08-26
GLM-5.2live since 2026-06-16
GLM-5.5watching
GLM-6watching
sweeping every 60 seconds

Know the minute Zhipu AI's next model drops. Free.

A free Chrome push the second it's confirmed on a public API, email about 15 minutes later. Texts or a phone call on paid plans.

  • Free forever
  • No card
  • Unsubscribe in one click

Also: the wire · the bench index · our track record