GLM-5.3

GLM-5.3 is Zhipu's current coding flagship. It is text-only, 1M-token context, 128K max output, and reasoning always on (low / high / max; default max). thinking.type: disabled fails. docs.z.ai/guides/overview/pricing.md still prices GLM-5.2 at $1.4 / $0.26 cached / $4.4 per 1M tokens. There is no GLM-5.3 token row. Coding Plan seats on the last first-party check of z.ai/subscribe are Lite $10 / Pro $30 / Max $60 per month, not a single $18 starter. Do not keep the vendor +50% / Terminal Bench 28.3 / "open in two weeks" numbers unless you re-open a first-party scorecard that still prints them.

Compare GLM-5.2 if you needed a live token price today, GLM Coding Plan for the seat table, and Claude Code if you wanted Anthropic's own agent.

Key Features

• Coding Plan first: Docs send you to Individual / Team Coding Plan, with off-peak and weekend calls at 50% points.
• Always-on reasoning: Migrate off thinking.type: disabled before you switch the model ID to glm-5.3.
• 1M / 128K: Context and max output from the model page. Confirm before you promise a bigger window.
• API later: OpenAI chat, Responses, and Anthropic message base URLs are listed. Availability is "soon," not a live $1.40 meter.

Use Cases

• People already on Lite $10 or Pro $30 who want the newest GLM coding checkpoint in Claude Code / Cline.
• Teams that need a public token price today: stay on GLM-5.2's $1.4 / $4.4 row.
• Self-host shoppers: weights were "planned" at launch. Do not treat MIT weights as shipped until a repo exists.

Limitation: Internal +50%, Claude Fable 5 comparisons, CyberGym 84.5%, and "MIT in two weeks" are vendor launch claims. We did not re-measure them.

Pricing

Coding Plan from z.ai/subscribe (same table as the GLM Coding Plan refresh). Token table from pricing.md.

If a leftover blog still says Coding Plan starts around $18 or copies GLM-5.2's $1.40 onto 5.3, treat these two pages as the source.

Getting Started

1. Open docs.z.ai/guides/llm/glm-5.3 and read the always-on reasoning note.
2. Buy a seat at z.ai/subscribe only after you know Lite $10 vs Pro $30.
3. Point your coding agent at the Coding Plan base URL, not a guessed token price.
4. Confirm GLM-5.2 if you needed pay-as-you-go today.

Frequently Asked Questions

Is GLM-5.3 $1.40 / $4.40?

Not on the public token table. That row is still GLM-5.2.

Is it open source today?

Do not mark it MIT-available.

Same as Coding Plan?

Coding Plan is the seat. GLM-5.3 is the model currently attached to that seat.

Alternatives

• GLM-5.2: Live token price.
• GLM Coding Plan: Lite $10 / Pro $30 / Max $60.
• Claude Code: Anthropic's first-party agent.

Tips

1. Quote Coding Plan $10 / $30 / $60, not $18.
2. Do not paste Terminal Bench 28.3 into a deck without a live scorecard.
3. Confirm the API banner before you write a 5.3 token integration.

Conclusion

GLM-5.3 is a Coding Plan model with no public token row yet, not an $18 unlimited or a copied $1.40 API. Start at the model page, then decide whether GLM-5.2 already covers the meter you need.