GLM-5.3 Max vs GPT-5.6 Terra Max: latency, speed and cost

GPT-5.6 Terra Max leads the intelligence index 32.3 to 31.7. GLM-5.3 Max costs 52% less on blended price. Benchmarks, sample bills, speed and a verdict.

Metric	GLM-5.3 (Max)	GPT-5.6 Terra (Max)
Intelligence Index	31.7	32.3
Input price (per 1M tokens)	$1.40	$2.00
Output price (per 1M tokens)	$4.40	$12.00
Output speed (tokens/s)	74	116
Time to first token	1.7s	3.4 min
Release date	2026-08-18	2026-07-09

Metric

GLM-5.3 (Max)

GPT-5.6 Terra (Max)

Intelligence Index

31.7

32.3

Input price (per 1M tokens)

$1.40

$2.00

Output price (per 1M tokens)

$4.40

$12.00

Output speed (tokens/s)

116

Time to first token

1.7s

3.4 min

Release date

2026-08-18

2026-07-09

If you only looked at the intelligence index, GLM-5.3 (Max) and GPT-5.6 Terra (Max) would be hard to tell apart. Everything below is read from the benchmark, price and speed data in Franklin AI's catalogue; nothing here comes from a hands-on test. GLM-5.3 Max starts answering far sooner, with a time to first token of 1.7 seconds compared with 3.4 minutes for GPT-5.6 Terra Max.

Throughput and time to first token

Once a response is under way, GLM-5.3 Max produces roughly 74 tokens per second against 116 tokens per second for GPT-5.6 Terra Max. GPT-5.6 Terra Max is ahead by 57% on throughput. Before any text appears, GLM-5.3 Max takes 1.7 seconds and GPT-5.6 Terra Max 3.4 minutes. GPT-5.6 Terra Max writes faster once it starts, but GLM-5.3 Max's head start is large enough that GLM-5.3 Max still finishes first at any typical reply length; GPT-5.6 Terra Max only catches up beyond about 41,000 tokens. A first-token wait of that length suits background work better than live conversation.

Put together, a 2,000-token answer would take about 29 seconds from GLM-5.3 Max and 3.7 minutes from GPT-5.6 Terra Max, counting the first-token wait.

Generation gap

GLM-5.3 Max arrived on August 18, 2026, 40 days after GPT-5.6 Terra Max (July 9, 2026). They are effectively contemporaries.

Running costs

The list prices are $1.40 in and $4.40 out for GLM-5.3 Max, and $2.00 in and $12.00 out for GPT-5.6 Terra Max, per million tokens. GPT-5.6 Terra Max costs 2.7 times as much per output token.

For a sample workload of 1 million input tokens and 200,000 output tokens, GLM-5.3 Max would cost $2.28 and GPT-5.6 Terra Max $4.40, a difference of $2.12. For writing 2,000 drafts of 2,000 output tokens from 300-token prompts, GLM-5.3 Max would come in 63% lower: $18.44 for GLM-5.3 Max against $49.20 for GPT-5.6 Terra Max. Blended at three input tokens per output token, the effective rates are $2.15 for GLM-5.3 Max and $4.50 for GPT-5.6 Terra Max.

The headline benchmark

GPT-5.6 Terra Max comes out ahead on the intelligence index, 32.3 to 31.7. On an aggregate like this, under a point is effectively level.

Across the 109 active models with an intelligence score in the catalogue, GLM-5.3 Max ranks 47th and GPT-5.6 Terra Max ranks 46th.

The catalogue does not yet split either model's score into coding and math indexes.

Is the price in proportion?

Per point of intelligence index, GLM-5.3 Max costs $0.068 of blended price per million tokens and GPT-5.6 Terra Max $0.14. GPT-5.6 Terra Max costs 2.1 times as much per point, which is the price of its lead.

Frequently asked questions

Is GLM-5.3 Max better than GPT-5.6 Terra Max?

Not across the board: GPT-5.6 Terra Max scores higher on the intelligence index, 32.3 to 31.7; GLM-5.3 Max is 52% cheaper on blended price; GPT-5.6 Terra Max streams output 1.6 times as fast; and GLM-5.3 Max reaches its first token sooner.

When were GLM-5.3 Max and GPT-5.6 Terra Max released?

GLM-5.3 Max is newer. It was released on August 18, 2026, 40 days after GPT-5.6 Terra Max (July 9, 2026).

Which is cheaper, GLM-5.3 Max or GPT-5.6 Terra Max?

GLM-5.3 Max is cheaper on a blended basis. GLM-5.3 Max lists $1.40 per million input tokens and $4.40 per million output tokens; GPT-5.6 Terra Max lists $2.00 and $12.00. For 1 million input and 200,000 output tokens that is $2.28 against $4.40.

Where do these numbers come from?

The figures are the recorded benchmark, price and speed data in Franklin AI's model catalogue. This page is written from that data automatically; neither model was tested by hand for it.

Does GLM-5.3 Max respond faster than GPT-5.6 Terra Max?

GPT-5.6 Terra Max generates output faster, at 116 tokens per second against 74 tokens per second. GLM-5.3 Max reaches its first token sooner (1.7 seconds against 3.4 minutes). Counting the first-token wait, GLM-5.3 Max finishes first for any reply shorter than about 41,000 tokens.

Final whistle

Bottom line: on balance the numbers favour GLM-5.3 Max, though GPT-5.6 Terra Max keeps a case of its own. Pick GPT-5.6 Terra Max if you generate very long outputs, beyond about 41,000 tokens. Pick GLM-5.3 Max if users are waiting on the first word and cost matters ($4.40 against $12.00 per million output tokens).

Where the numbers come from

The scores, speed and prices in this comparison come from independent third-party model evaluations that Franklin AI syncs each day, and the article was written from a snapshot of them. For the latest results, open the GLM-5.3 (Max) and GPT-5.6 Terra (Max) pages. The AI Benchmark methodology explains how that data is collected and shown.