GLM 5.3 Released: API Pricing & GLM 5.5 Status

Here’s the short answer: Zhipu (Z.ai) released GLM 5.3 on August 14, 2026, followed by the lighter GLM 5.3 Flash later in August. Both are already available on APIMart, where you can try GLM 5.3 with one API key. GLM 5.5 has not been released yet: as of October 2026 there is no official model card, weights, API ID, or price.

If you only need the takeaway:

• Use GLM 5.3 for coding, debugging, and agent work where quality matters

• Use GLM 5.3 Flash for high-volume, low-cost jobs like classification, extraction, and routine chat

• Don’t wait for GLM 5.5 to ship a product; there is no confirmed date

Key Takeaways

• GLM 5.3 launched on August 14, 2026, first through Zhipu’s GLM Coding Plan. Zhipu positions it as a coding-focused model.
• Zhipu says GLM 5.3 shares the same base model as GLM 5.2 and gets its gains from extended post-training, not a new pretraining run.
• Zhipu claims a roughly 50% improvement in coding capability over GLM 5.2. These are Zhipu’s own evaluations, not independent results.
• GLM 5.3 on APIMart costs $1.12 per 1M input tokens and $3.52 per 1M output tokens; GLM 5.3 Flash costs $0.06 and $0.20.
• GLM 5.5 is still unreleased. Analyst forecasts pointed to August 2026, but that window passed without an official launch.
• APIMart serves every GLM model from one OpenAI-compatible endpoint, https://api.apimart.ai/v1, so switching versions only means changing the model value.

Quick Status Check

1. What GLM 5.3 Actually Changed

GLM 5.3 is an upgrade of GLM 5.2 rather than a brand-new model. According to Zhipu, it uses the same base model and improves through longer post-training focused on code.

What Zhipu published at launch:

• Coding first. The model shipped through the GLM Coding Plan and works with coding agents such as Claude Code, OpenCode, and Zhipu’s ZCode.

• About 50% better at coding than GLM 5.2, by Zhipu’s internal measurement.

• Security work. Zhipu reports that GLM 5.3 helped security teams find 2,436 vulnerabilities across 269 open-source projects, and it published cybersecurity benchmark results such as CyberGym 84.5% and ExploitBench 54.4% (up from 24.4% for its predecessor).

• Open weights later. Zhipu said the weights would follow after safety evaluation and hardening.

Keep one caveat in mind: almost every number above comes from Zhipu itself. Independent reviewers have pointed out that GLM 5.3 still trails top closed models on several of the same benchmarks. Test it on your own prompts before switching production traffic.

2. GLM 5.3 vs GLM 5.3 Flash: Which One to Use

The two models split by cost and depth, not by modality.

A common setup is to route easy steps to GLM 5.3 Flash and escalate only the hard steps to GLM 5.3. Because both share the same API format, that routing is a one-line change.

3. GLM API Pricing on APIMart

These are current APIMart prices in USD per 1M tokens. Check the GLM API pricing page for live rates, cached-input pricing, and member discounts.

Two things stand out:

• GLM 5.3 costs about 22% more on input and 10% more on output than GLM 5.2, so if its coding gains hold up on your tasks, the upgrade is cheap.

• GLM 5.3 Flash is roughly 18× cheaper on output than GLM 5.3, which makes it the obvious default for bulk work.

4. Using GLM 5.3 Through APIMart

APIMart exposes GLM 5.3, GLM 5.3 Flash, and older GLM versions through one endpoint: https://api.apimart.ai/v1. The request format is OpenAI-compatible, so the same SDKs work across Python, Node.js, and Java. To switch versions, change the model value to glm-5.3, glm-5.3-flash, or glm-5.2.

On APIMart, GLM 5.3 accepts up to 128K input tokens and returns up to 16K output tokens per request. Billing is pay-as-you-go in USD, and APIMart maintains a 99.9% uptime SLA with automatic failover ^[1][2].

You can also pair GLM with other models on the same key, for example GLM 5.3 for planning and analysis, then a video or image model for generation, under one balance.

5. What About GLM 5.5?

GLM 5.5 has not been released. A JPMorgan research note reported in June 2026 pointed to an August launch, and later rumors moved it to September, but neither window produced an official announcement, model card, weights, API ID, or price.

Until Zhipu publishes real specs, treat any GLM 5.5 parameter counts, context sizes, or prices you see online as rumors. Our advice: build on GLM 5.3 now. When GLM 5.5 ships and lands on APIMart, moving to it is the same one-line model change.

FAQs

Is GLM 5.3 released?

Yes. Zhipu released GLM 5.3 on August 14, 2026, and the lighter GLM 5.3 Flash followed later in August. Both are available on APIMart as glm-5.3 and glm-5.3-flash.

How much does the GLM 5.3 API cost?

On APIMart, GLM 5.3 costs $1.12 per 1M input tokens and $3.52 per 1M output tokens. GLM 5.3 Flash costs $0.06 and $0.20. See the GLM API page for live prices and member discounts.

Is GLM 5.5 out yet?

No. As of October 2026, Zhipu has not released GLM 5.5 or published an official date. Forecasts for August and September 2026 passed without a launch.

Should I use GLM 5.3 or GLM 5.3 Flash?

Use GLM 5.3 for coding, debugging, and agent steps where accuracy matters. Use GLM 5.3 Flash for high-volume, cost-sensitive tasks such as classification, extraction, and routine chat. Many teams route between the two by step.