Agenci AI · karta decyzji
GLM-5.2
Z.ai
GLM-5.2 is an open-source, large-scale LLM with advanced reasoning, text generation, and contextual understanding, designed for code and long-form tasks with a 1-million-token context window.
Dostęp
Open source
Cena od
18
Trial
Free access to chat.z.ai for testing; API and Coding Plan require payment.
Sprawdzone
9.08.2026
W skrócie
Czy to narzędzie pasuje do Twojego zadania?
Free access via chat.z.ai; paid plans include a $18/month Coding Plan and API pricing at $1.40 per million input tokens, $4.40 per million output tokens, and $0.26 per million cache tokens.
Zastosowania
- Refactoring entire code repositories
- Automated research and test benches
- Debugging across dozens of files
- Creating complete prototypes or mini-games from a single prompt
Najważniejsze funkcjonalności
- context window of up to 1 million tokens
- multi-step reasoning capabilities
- code generation and analysis
- support for agent-based workflows
- tool calling
- structured output
- support for reasoning effort adjustment
- available via Hugging Face and Ollama
- supports deployment with multiple frameworks
Darmowy plan i trial
Darmowy dostęp: GLM-5.2 is available for free via the chat.z.ai interface and its weights can be downloaded under the MIT license.
Trial: Free access to chat.z.ai for testing; API and Coding Plan require payment.
Cennik: Free access via chat.z.ai; paid plans include a $18/month Coding Plan and API pricing at $1.40 per million input tokens, $4.40 per million output tokens, and $0.26 per million cache tokens.
Plany i limity
| Plan | Cena | Dostęp | Limit / zawartość | Trial / karta |
|---|---|---|---|---|
| Free Chat Access | 0 USD / free | free | chat.z.ai: questions, trials, small sites | unknown · karta: no |
| GLM Coding Plan | 18 USD / monthly | subscription | GLM Coding Plan: agents for code | unknown · karta: yes |
| API Usage | 1.40 USD / per million tokens | pay_per_use | API for applications and agents in production | unknown · karta: yes |
Dane i bezpieczeństwo
- Prywatność
- unknown
- Trening na danych
- unknown
- Region danych
- unknown
- RODO / GDPR
- unknown
- Bezpieczeństwo
- unknown
Mocne strony
- + One million token context window that remains stable
- + Open weights under MIT license
- + Adjustable reasoning effort levels
- + Available through 20+ hosting providers
- + Designed for long-running agent tasks and code projects
Na co uważać
- Input and output limited to text only
- Weights too large for personal GPUs
- Quotas on Coding Plan may be restrictive during peak hours