GLM 5.3 and GLM 5.3 Flash from Z.ai (Zhipu AI) are among the most visited model pages on LLM Perks this month. The reason is simple: they are open-weight models with a 1M-token context window, strong coding benchmarks, and an unusual number of free ways to try them. This guide explains the difference between the two models, lists every free route we track, shows the official API price list, and covers how GLM fits into coding tools and subscriptions. Pricing comes from Z.ai's documentation; availability comes from our catalog, last refreshed November 12, 2026.
GLM 5.3 versus GLM 5.3 Flash
Z.ai sells the latest generation in three variants: GLM-5.3 (the flagship), GLM-5.3-Flash (the fast, low-cost model), and GLM-5.3-FlashX (a faster tier of Flash). Our catalog tracks GLM 5.3 and GLM 5.3 Flash, plus a Flash Thinking variant, and lists benchmark data from Artificial Analysis.
| Metric (Artificial Analysis) | GLM 5.3 | GLM 5.3 Flash |
|---|---|---|
| Intelligence Index | 44.8 | 41.8 |
| GPQA Diamond | 91.7% | 91.2% |
| Humanity's Last Exam | 42.3% | 39.9% |
| Long-context reasoning (LCR) | 79.7% | 80.0% |
| Context window | 1M tokens | 1M tokens |
| Weights | Open | Open |
| Size class in our catalog | Frontier | Mid-size |
The gap on the headline index is about three points, while on long-context reasoning the two are effectively level. For many coding and agent tasks, Flash is the practical choice: it costs a fraction of the flagship and has more free routes. Choose the flagship when the harder reasoning benchmarks matter for your workload, and test both on your own tasks because leaderboard rankings are a weak predictor of how a model behaves in your repository.
Where GLM 5.3 is free
As of November 12, 2026, our catalog lists one free route for the flagship and four for Flash.
GLM 5.3 (flagship)
| Provider | Model ID | Context | Terms |
|---|---|---|---|
| Together AI | zai-org/GLM-5.3-FP8 | 1M | Fair use, depends on server load |
| Together AI | zai-org/GLM-5.3-FP8-Lora | 1M | Fair use, depends on server load |
"Fair use" means there is no published fixed quota, and throughput varies with load. That is fine for experiments and unattended batch jobs that can retry, and less suitable for latency-sensitive production traffic.
GLM 5.3 Flash
| Provider | Model ID | Terms |
|---|---|---|
| OrcaRouter | z-ai/glm-5.3-flash-free | Free tier |
| NVIDIA Build | z-ai/glm-5.3-flash | Free serverless inference for development; no card; limits may change |
| UnoRouter | glm-5.3-flash:free | No-card free models share capacity and have per-model limits; 1M context |
| OpenCode Zen | glm-5.3-flash-free | Free for a limited time |
The catalog shows GLM 5.3 Flash as free since September 19, 2026. OpenCode's documentation says GLM-5.3-Flash Free is free for a limited time and that its provider follows a zero-retention policy and does not use your data for model training.
The official API price list
When you outgrow the free routes, Z.ai's own API is inexpensive. These are the prices per 1M tokens from the Z.ai pricing page.
| Model | Input | Cached input | Cached input storage | Output |
|---|---|---|---|---|
| GLM-5.3 | $1.40 | $0.26 | Limited-time free | $4.40 |
| GLM-5.3-Flash | $0.15 | $0.03 | Limited-time free | $0.50 |
| GLM-5.3-FlashX | $0.37 | $0.075 | Limited-time free | $1.25 |
| GLM-5.2 | $1.40 | $0.26 | Limited-time free | $4.40 |
| GLM-5 | $1.00 | $0.20 | Limited-time free | $3.20 |
| GLM-4.7 | $0.60 | $0.11 | Limited-time free | $2.20 |
Z.ai also keeps a small set of older models permanently free on its own API: GLM-4.7-Flash, GLM-4.5-Flash, and GLM-4.6V-Flash (a vision model) are listed at no charge for input, output, and cached input. They are older and weaker than the 5.3 generation, but they are a stable option when you want a free model with a first-party SLA-less endpoint and no reliance on a third-party gateway. Our credits page tracks that offer.
What Flash costs in practice
Take a task with 1M input tokens (no caching) and 200K output tokens:
- GLM-5.3-Flash: 1M × $0.15 + 0.2M × $0.50 = $0.15 + $0.10 = $0.25
- GLM-5.3: 1M × $1.40 + 0.2M × $4.40 = $1.40 + $0.88 = $2.28
The flagship is roughly nine times more expensive for the same token volume. Caching narrows the gap, because cached input costs a small fraction of the normal rate, so repeated repository context is cheap on either model.
Using GLM 5.3 in coding tools
Z.ai's documentation describes using GLM models for AI coding in Claude Code, Kilo Code, Cline, OpenCode, and OpenClaw, and sells subscription plans designed for that use.
The GLM Coding Plan
Z.ai's Coding Plan is a flat monthly subscription for IDE and agent use, with plans starting at $18/month on the Z.ai pricing material. Our Z.ai subscription page compares the Lite, Pro, and Max tiers against pay-as-you-go API prices so you can see where each pays off. If you use a coding agent for hours every day, a flat plan usually beats metered pricing; if your use is occasional, metered Flash is cheaper.
OpenCode Go
OpenCode Go lists GLM-5.3-Flash, GLM-5.3, and GLM-5.2 among its models. Go costs $10/month and Go Plus $40/month. OpenCode's docs describe the allowance in dollar terms: each plan has a five-hour limit equal to 20% of the monthly limit, a weekly limit equal to 50%, and the monthly limit itself. Read our explainer on OpenCode Zen and Go for details.
An OpenAI-compatible call
from openai import OpenAI
client = OpenAI(
base_url="https://api.together.xyz/v1",
api_key="YOUR_TOGETHER_API_KEY",
)
reply = client.chat.completions.create(
model="zai-org/GLM-5.3-FP8",
messages=[{"role": "user", "content": "Refactor this function for readability."}],
)
print(reply.choices[0].message.content)Swap the base URL and model ID to use any other route from the tables above. All of them speak the OpenAI chat-completions format.
How to choose
| You want | Pick |
|---|---|
| The strongest GLM, free, for experiments | GLM 5.3 on Together AI (fair use) |
| The cheapest capable coding model | GLM 5.3 Flash, free routes or $0.15/$0.50 per 1M tokens |
| No card at all | NVIDIA Build or UnoRouter free models |
| A free model with stable terms | GLM-4.7-Flash on Z.ai (older, weaker) |
| All-day agent use at a flat price | GLM Coding Plan or OpenCode Go |
Caveats
- "Free" is a promotion on most routes. OpenCode says its free models are available for a limited time, and Z.ai's cache storage is "limited-time free."
- Privacy differs. Read each provider's data policy. NVIDIA's free endpoints are for trial use and log usage; OpenCode's GLM-5.3-Flash route states zero retention.
- Open weights do not mean easy local use. The flagship is a very large model. For local runs see our guide to running LLMs locally.
- Benchmarks are third-party. Use them to shortlist, then test.
FAQ
Is GLM 5.3 free?
Yes, at Together AI on a fair-use basis, as of November 12, 2026. GLM 5.3 Flash is free at four providers. Z.ai's own API charges the prices in the table.
What is the difference between GLM-5.3-Flash and FlashX?
Both are Flash-class models. Z.ai prices FlashX higher ($0.37 input and $1.25 output per 1M tokens versus $0.15 and $0.50), which corresponds to a faster tier.
Does GLM 5.3 have a 1M context window?
Our catalog lists up to 1M tokens of context for both models on the routes that expose it. Check the row for your provider, because routes can cap it lower.
Which free GLM route needs no credit card?
NVIDIA Build and UnoRouter's no-card free models both list no card requirement. OpenCode Zen and OrcaRouter state their own terms on the model page.
Can I use GLM in Claude Code or Cline?
Z.ai's documentation lists Claude Code, Kilo Code, Cline, OpenCode, and OpenClaw among supported tools. Follow Z.ai's guide for the current settings.
Where can I see every free model?
Open the catalog and filter for free models, or browse the GLM family page.
