DeepSeek V4 Pro 0813 Max vs GPT-5.6 Luna Max compared

GPT-5.6 Luna Max leads the intelligence index 23.8 to 23.1. GPT-5.6 Luna Max costs 77% less on blended price. Benchmarks, sample bills, speed and a verdict.

Metric	DeepSeek V4 Pro 0813 (Max)	GPT-5.6 Luna (Max)
Intelligence Index	23.1	23.8
Input price (per 1M tokens)	$1.32	$0.20
Output price (per 1M tokens)	$3.96	$1.20
Output speed (tokens/s)	67	125
Time to first token	1.7s	108s
Release date	2026-08-13	2026-07-09

Metric

DeepSeek V4 Pro 0813 (Max)

GPT-5.6 Luna (Max)

Intelligence Index

23.1

23.8

Input price (per 1M tokens)

$1.32

$0.20

Output price (per 1M tokens)

$3.96

$1.20

Output speed (tokens/s)

125

Time to first token

1.7s

108s

Release date

2026-08-13

2026-07-09

If you only looked at the intelligence index, DeepSeek V4 Pro 0813 (Max) and GPT-5.6 Luna (Max) would be hard to tell apart. Time to first token is the standout number here: 1.7 seconds for DeepSeek V4 Pro 0813 Max, 108 seconds for GPT-5.6 Luna Max. The figures come from tracked catalogue data, not from our own testing, so treat them as a starting point for a trial on your tasks.

GPT-5.6 Luna Max is the faster writer

DeepSeek V4 Pro 0813 Max generates about 67 tokens per second and GPT-5.6 Luna Max about 125 tokens per second. GPT-5.6 Luna Max is ahead by 86% on throughput. The wait for the first token averages 1.7 seconds on DeepSeek V4 Pro 0813 Max against 108 seconds on GPT-5.6 Luna Max. The faster writer, GPT-5.6 Luna Max, gives away too much at the start to win on time: counting the first-token wait, DeepSeek V4 Pro 0813 Max finishes first for any reply shorter than about 15,000 tokens. A first-token wait of that length suits background work better than live conversation.

For a 600-token response, first token plus generation comes to roughly 11 seconds on DeepSeek V4 Pro 0813 Max and 113 seconds on GPT-5.6 Luna Max.

How the intelligence index compares

DeepSeek V4 Pro 0813 Max records 23.1 on the intelligence index and GPT-5.6 Luna Max records 23.8. On an aggregate like this, under a point is effectively level.

For context, that places DeepSeek V4 Pro 0813 Max 72nd and GPT-5.6 Luna Max 70th out of 109 scored models still on offer.

For coding or maths work specifically, the catalogue has no dedicated index on either side yet.

Price per token

The list prices are $1.32 in and $3.96 out for DeepSeek V4 Pro 0813 Max, and $0.20 in and $1.20 out for GPT-5.6 Luna Max, per million tokens. DeepSeek V4 Pro 0813 Max costs 3.3 times as much per output token.

For a sample workload of 1 million input tokens and 200,000 output tokens, DeepSeek V4 Pro 0813 Max would cost $2.11 and GPT-5.6 Luna Max $0.44, a difference of $1.67. For writing 2,000 drafts of 2,000 output tokens from 300-token prompts, GPT-5.6 Luna Max would come in 70% lower: $16.63 for DeepSeek V4 Pro 0813 Max against $4.92 for GPT-5.6 Luna Max. On a 3:1 blend of input to output tokens, that works out to $1.98 versus $0.45 per million.

Is the price in proportion?

Dividing the blended price by the intelligence index gives a rough cost per point: $0.086 for DeepSeek V4 Pro 0813 Max and $0.019 for GPT-5.6 Luna Max, per million tokens. GPT-5.6 Luna Max scores higher and costs less per point, so on price per point there is nothing to trade off.

Generation gap

GPT-5.6 Luna Max came first, on July 9, 2026; DeepSeek V4 Pro 0813 Max followed on August 13, 2026. They are effectively contemporaries.

Frequently asked questions

Is GPT-5.6 Luna Max better than DeepSeek V4 Pro 0813 Max?

Neither wins everywhere: GPT-5.6 Luna Max scores higher on the intelligence index, 23.8 to 23.1; GPT-5.6 Luna Max is 77% cheaper on blended price; GPT-5.6 Luna Max streams output 1.9 times as fast; and DeepSeek V4 Pro 0813 Max reaches its first token sooner.

Which model has lower latency?

GPT-5.6 Luna Max generates output faster, at 125 tokens per second against 67 tokens per second. DeepSeek V4 Pro 0813 Max reaches its first token sooner (1.7 seconds against 108 seconds). Counting the first-token wait, DeepSeek V4 Pro 0813 Max finishes first for any reply shorter than about 15,000 tokens.

Who makes DeepSeek V4 Pro 0813 Max and GPT-5.6 Luna Max?

DeepSeek V4 Pro 0813 Max is made by DeepSeek and GPT-5.6 Luna Max by OpenAI. DeepSeek V4 Pro 0813 Max was released on August 13, 2026 and GPT-5.6 Luna Max on July 9, 2026.

How much do DeepSeek V4 Pro 0813 Max and GPT-5.6 Luna Max cost?

GPT-5.6 Luna Max is cheaper on a blended basis. DeepSeek V4 Pro 0813 Max lists $1.32 per million input tokens and $3.96 per million output tokens; GPT-5.6 Luna Max lists $0.20 and $1.20. For 1 million input and 200,000 output tokens that is $2.11 against $0.44.

When were DeepSeek V4 Pro 0813 Max and GPT-5.6 Luna Max released?

DeepSeek V4 Pro 0813 Max is newer. It was released on August 13, 2026, 35 days after GPT-5.6 Luna Max (July 9, 2026).

Final whistle

In one line: on balance the numbers favour DeepSeek V4 Pro 0813 Max, though GPT-5.6 Luna Max keeps a case of its own. Pick DeepSeek V4 Pro 0813 Max if users are waiting on the first word and replies need to finish sooner at typical lengths. Pick GPT-5.6 Luna Max if cost matters ($1.20 against $3.96 per million output tokens) and you generate very long outputs, beyond about 15,000 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 DeepSeek V4 Pro 0813 (Max) and GPT-5.6 Luna (Max) pages. The AI Benchmark methodology explains how that data is collected and shown.