GPT 2.6, Veo 3.1, Seedance 2.5, Kling 4 in deAPI!
Web-search and web-contents are now available via API
Browse all Models
Browse all
Browse all
Browse all

LLM

GLM-5.3 API, free to try

GLM-5.3 is an open-weight large language model from Z.ai. It reads and writes text, holds up to 1M tokens of context and has three reasoning levels. deAPI serves it on an OpenAI-compatible endpoint, so the OpenAI SDK works once you change the base URL, the key and the model ID.

  • Text in, text out, a 1,000,000-token context window
  • json_object output, tool calls and streaming
  • reasoning_effort in three levels: low, high and max

No credit card · Free credits valid for 30 days · Then top up and pay as you go, no subscription

GLM-5.3 LLM

model: "glm-5.3"
reasoning_effort: "high"

system: You are an on-call SRE. Name the root cause in at most two sentences, then give a one-line fix. Plain text, no Markdown.

user

Logins return 500 since 09:12. What is the root cause?

09:12:01 INFO  api   GET /reports/418 200 41ms
09:12:07 WARN  redis OOM command not allowed when used memory > 'maxmemory'
09:12:07 ERROR api   POST /login 500 SessionStoreError: could not save session
09:12:08 ERROR api   POST /login 500 SessionStoreError: could not save session
09:12:09 INFO  api   GET /health 200 2ms
assistant

Root cause: Redis hit its maxmemory limit at 09:12:07 and started rejecting writes (OOM command not allowed), so the login endpoint can no longer save sessions, causing 500 errors.

Fix: Free up or raise Redis memory (e.g., flush stale keys / increase maxmemory or set an eviction policy like allkeys-lru), then restart or re-enable writes.

168 input + 79 output tokens (0 reasoning) · 2.2 s Cost on deAPI: $0.00064 (price as of November 2026)

GLM-5.3 use cases

Four jobs where the model fits, each with the number behind it. All of them run on the same deAPI key as transcription, OCR and image generation.

  1. Logs and stack traces. Paste the lines around an error and ask for the root cause. In three runs of the question above, all three answers named the Redis maxmemory limit, not the 500 on /login. Run 1 used 168 input and 79 output tokens.

  2. Coding agents. Z.ai reports 88.2 on Terminal Bench 2.1 (Claude Code harness) and 66.9 on DeepSWE v1.1 (mini-swe-agent harness). Coding tools that speak the OpenAI Chat Completions API can point at https://oai.deapi.ai/v1.

  3. Long transcripts and documents. The context window holds 1,000,000 tokens and one answer can run to 131,072. Transcribe a day of sales calls with WhisperLargeV3Ct2 and ask for objections, follow-ups and quotes in one request.

  4. Agents that call tools. Given a get_logs tool and the request “Logins fail with 500. Check the last 50 log lines of the api service.”, the model returned a tool call with {"lines": 50, "service": "api"} (one run). Your code runs the call and sends the result back.

GLM-5.3 API pricing

You pay per token from the same prepaid balance as every other deAPI model. Reasoning tokens are part of completion_tokens in usage. No subscription.

Input (text) $1.54 ≈ 3.2M tokens
Cached input (repeated prompt prefix) $0.308 ≈ 16M tokens
Output (answer and reasoning) $4.84 ≈ 1.0M tokens
Log question in the hero, one call $0.00064 ≈ 7,700 calls
Log line to JSON (Python example), one call $0.00025 ≈ 19,000 calls

What the free account includes

  • $5 of credits once you add a payment card
  • About 7,700 calls like the log question in the hero
  • Credits valid for 30 days, then top up and pay as you go
Get $5 credits

Prices for glm-5.3 in USD, November 2026. Token counts from the hero question and the Python example. Context and output limits from GET https://oai.deapi.ai/v1/models.

GLM-5.3 vs GLM-5.3 Flash vs Kimi K3

Input Text Text, image Text, image
Context 1,000,000 tokens 1,048,576 tokens 1,000,000 tokens
Max output 131,072 tokens 131,072 tokens 131,072 tokens
reasoning_effort low, high, max none to max, 7 values; reasons even at none none to max, 7 levels
response_format json_object json_schema, json_object json_schema, json_object
Price per 1M, input / output $1.54 / $4.84 $0.165 / $0.55 $3.30 / $16.50

Limits and input types from GET /v1/models. Accepted reasoning_effort and response_format values from test calls on 16 November 2026.

GLM-5.3 API example

One POST to https://oai.deapi.ai/v1/chat/completions with your deAPI key as the Bearer token. In the OpenAI SDK, set base_url, your deAPI key and the model ID; the rest of a Chat Completions call stays the same. Replace YOUR_API_KEY with a key from your dashboard.

Model ID
glm-5.3
Endpoint
POST https://oai.deapi.ai/v1/chat/completions
Input
Text only. A request with an image_url part returns a 400 error.
Context and output
1,000,000-token context window, up to 131,072 tokens per answer
Reasoning
reasoning_effort: low, high or max. Other values return a 400 error. When the model reasons, the text comes back in reasoning_content.
Output options
response_format with json_object, tools, stream. json_schema is not supported, so name the keys in the prompt.
curl https://oai.deapi.ai/v1/chat/completions \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "glm-5.3",
    "reasoning_effort": "low",
    "messages": [
      {"role": "system", "content": "Sort support tickets. Reply with one word: billing, bug, feature or other."},
      {"role": "user", "content": "I was charged twice for the November top-up."}
    ]
  }'
import json
from openai import OpenAI

client = OpenAI(base_url="https://oai.deapi.ai/v1", api_key="YOUR_API_KEY")

log_line = "09:12:07 ERROR api POST /login 500 SessionStoreError: could not save session"

response = client.chat.completions.create(
    model="glm-5.3",
    reasoning_effort="low",
    messages=[
        {"role": "system", "content": "Read the log line. Reply with a JSON object "
         "with exactly these keys: service (string), level (string), "
         "status (number), error (string)."},
        {"role": "user", "content": log_line},
    ],
    response_format={"type": "json_object"},
)

entry = json.loads(response.choices[0].message.content)
print(entry["service"], entry["level"], entry["status"])  # api ERROR 500
print(entry["error"])  # SessionStoreError: could not save session
print(response.usage)

What is GLM-5.3?

GLM-5.3 is an open-weight text model from Z.ai, published on Hugging Face under the GLM-5.3 License. Z.ai’s model card says it uses the same base model as GLM-5.2 and that every gain comes from post-training. It takes text in, writes text out and has three reasoning levels. On deAPI you call it like any OpenAI chat model and pay per token.

  1. Create a free account

    Sign up and get $5 of credits.

  2. Point your client at deAPI

    Set base_url to https://oai.deapi.ai/v1 and use your deAPI key.

  3. Send the request

    Use glm-5.3 as the model and read the answer from choices[0].message.content.

GLM-5.3 FAQ

Yes. deAPI serves it at https://oai.deapi.ai/v1 under the model ID glm-5.3, with the request and response format of the OpenAI Chat Completions API. In code written for the OpenAI SDK you change three values: base_url, the API key and the model. Run your own tests before you switch traffic.
A new deAPI account gets $5 of credits once you add a payment card, enough for about 7,700 calls like the log question on this page. The credits are valid for 30 days. After that you top up and pay as you go, with no subscription.
$1.54 per million input tokens and $4.84 per million output tokens (November 2026). Reasoning tokens count as output, and you see them in usage.completion_tokens_details.reasoning_tokens.
No. GLM-5.3 on deAPI takes text only, and a request with an image_url part returns a 400 error. For images, call glm-5.3-flash on the same endpoint. Run PDFs and scans through the OCR API and send the text.
On deAPI the context window is 1,000,000 tokens and one answer can be up to 131,072 tokens long. Reasoning tokens count toward max_tokens.
Z.ai’s model card lists 88.2 on Terminal Bench 2.1, 66.9 on DeepSWE v1.1, 28.3 on Terminal Bench 3.0 and 84.5 on CyberGym, the last two at max reasoning effort. These are Z.ai’s numbers; test it on your own data with the free credits.
reasoning_effort takes low, high or max; there is no none. On the ticket request from the cURL example, low answered “billing” with 2 output tokens and high gave the same answer with 5 output tokens, 3 of them reasoning (one run each). For JSON, use response_format: {"type": "json_object"} and name the keys in the prompt, as in the Python example. json_schema returns an unsupported_parameter error.
Z.ai publishes the weights on Hugging Face under its own GLM-5.3 License. The license grants the right to use, modify, distribute and sell copies. Companies that run a model-as-a-service business with more than 10 billion US dollars of revenue in 12 months must pass a Z.ai security review first. Calls through deAPI are also covered by the deAPI Terms of Service.

Send your first GLM-5.3 request with $5 of free credits

Sign up with e-mail or Google. No subscription, pay as you go.