Model comparison
GLM-5.3 vs GPT-6 Astra
GPTCompare published specifications, vendor API list prices and measured Borg performance. Both models are included in Borg.
GLM-5.3
GPT-6 Astra
Included in Borg
At a glance
| Specification | GLM-5.3 | GPT-6 Astra |
|---|---|---|
| Provider | Z.ai | OpenAI |
| Release | 2026-08-18 Vendor source | 2026-09-03 Vendor source |
| Context | 1,048,576 tokens Vendor source | 1,050,000 tokens Vendor source |
| Modalities | Input: text; output: text | Input: text, image; output: text |
| Reasoning / thinking modes | Yes / Not published | Yes / Not published |
| Open weights | Yes | No |
| Vendor API list price / 1M tokens | $1.4 input / $4.4 output | $10 input / $50 output Standard global rates for prompts up to 272,000 input tokens. Above 272,000 input tokens, input and cached-input rates are 2× and output rates are 1.5× for the full request. |
| Borg | Included in Borg subscription | Included in Borg subscription |
Z.aiMeasured on Borg
Metrics are not directly comparable to vendor benchmark claims. Models served different requests; this is observed gateway traffic, not a controlled head-to-head test.
Time to first token (gateway proxy)
Output tokens per second (whole turn)
| Metric | GLM-5.3 | GPT-6 Astra |
|---|---|---|
| Time to first token (gateway proxy) p50 | 7293 ms (27495 samples) | 4780 ms (1335 samples) |
| Tokens per second (end to end) p50 | 25.2 output tokens per second (626315 samples) | 16 output tokens per second (18243 samples) |
| Request success rate | Not published | 99.6% (2222 samples) |
| Coding bench pass rate | 53% (15 samples) | 67% (15 samples) |
Measurement window, samples and methodology
The fleet reference is the median of published model medians, not a request-weighted result. Prompt lengths, reasoning settings and traffic differ between models.
GLM-5.3: Measured on Borg: SearchAtlas AI Gateway production request logs. 30-day window ending 2026-10-08; 659807 requests. Minimum 100 samples per published request metric.
GPT-6 Astra: Measured on Borg: SearchAtlas AI Gateway production request logs. 30-day window ending 2026-10-08; 20969 requests. Minimum 100 samples per published request metric.
Completion speed covers the whole turn including queueing and reasoning time; it is not a streaming tokens-per-second figure.
- Time to first token (gateway proxy): time until the first streamed response event reaches the gateway; our closest available measure of time to first token
- Tokens per second (end to end): tokens per second, end to end: provider-reported output tokens divided by the whole turn's elapsed time, including queueing and reasoning; not a streaming per-token rate
How to choose
GLM-5.3 is listed for complex coding-agent tasks, whereas GPT-6 Astra is listed for hard coding assignments. Choosing between GLM-5.3 and GPT-6 Astra requires checking published specifications against observed Borg traffic for your workflow.
Key differences
| Spec | GLM-5.3 | GPT-6 Astra |
|---|---|---|
| Context window | 1,048,576 tokens Vendor source | 1,050,000 tokens Vendor source Larger context |
| Reasoning modes | Yes / Not published | Yes / Not published |
| Release date | 2026-08-18 Vendor source | 2026-09-03 Vendor source |
| API price input / 1M | 1.4 Lower listed rate | 10 Standard global rates for prompts up to 272,000 input tokens. Above 272,000 input tokens, input and cached-input rates are 2× and output rates are 1.5× for the full request. |
| API price output / 1M | 4.4 Lower listed rate | 50 Standard global rates for prompts up to 272,000 input tokens. Above 272,000 input tokens, input and cached-input rates are 2× and output rates are 1.5× for the full request. |
| Time to first token (gateway proxy) | 7293 ms (27495 samples) | 4780 ms (1335 samples) Lower median |
| Output tokens per second (whole turn) | 25.2 output tokens/s (whole turn) (626315 samples) Higher measured rate | 16 output tokens/s (whole turn) (18243 samples) |
| Request success | Not published | 99.6% (2222 samples) |
| Coding bench | 53% (15 samples) | 67% (15 samples) Higher pass rate |
FAQ
How do GLM-5.3 and GPT-6 Astra compare for coding?
GLM-5.3 has a Borg coding benchmark of 53.33% pass rate across 15 samples and 3 tasks; GPT-6 Astra has 66.67% pass rate across 15 samples and 3 tasks. The GLM-5.3–GPT-6 Astra coding comparison cannot establish an overall winner from this small test alone.
How quickly do GLM-5.3 and GPT-6 Astra start responding on Borg?
Borg median first response for GLM-5.3 is 7,293 ms; for GPT-6 Astra it is 4,780 ms. The GLM-5.3–GPT-6 Astra gateway medians are observations, not guaranteed latency for individual requests.
What are the API prices for GLM-5.3 and GPT-6 Astra?
The listed API rate for GLM-5.3 is $1.4 input / $4.4 output per 1M tokens; for GPT-6 Astra it is $10 input / $50 output per 1M tokens. Neither GLM-5.3 nor GPT-6 Astra has its Borg subscription charge represented by those provider API rates. GLM-5.3 vendor source · GPT-6 Astra vendor source
Claude