GPT-5.6 Luna Max vs Mistral Large 4 High compared

On the headline score, OpenAI's GPT-5.6 Luna (Max) and Mistral's Mistral Large 4 (High) are almost level. All numbers are the catalogue's recorded values; nobody ran either model by hand for this page. Time to first token is the standout number here: 1.7 seconds for Mistral Large 4 High, 108 seconds for GPT-5.6 Luna Max.

Mistral Large 4 High is the faster writer

Once a response is under way, GPT-5.6 Luna Max produces roughly 125 tokens per second against 139 tokens per second for Mistral Large 4 High. That makes Mistral Large 4 High roughly 1.1 times as fast at writing. The wait for the first token averages 108 seconds on GPT-5.6 Luna Max against 1.7 seconds on Mistral Large 4 High. GPT-5.6 Luna Max waits 62 times as long before its first token. Waits that long are felt on every chat-style request, though they matter less for batch jobs.

Add the two numbers up for a 2,000-token reply and Mistral Large 4 High finishes first: 2.1 minutes for GPT-5.6 Luna Max, 16 seconds for Mistral Large 4 High.

Mistral Large 4 High leads on the intelligence index

The catalogue puts GPT-5.6 Luna Max at 23.8 and Mistral Large 4 High at 26.1 on its intelligence index. That is a modest edge: real, but small enough that the right set of tasks could reverse it. Put another way, GPT-5.6 Luna Max would need 10% more on this index to draw level.

Across the 109 active models with an intelligence score in the catalogue, GPT-5.6 Luna Max ranks 70th and Mistral Large 4 High ranks 65th.

Specialist coding and math indexes are missing for both, so treat the overall score as the only capability signal here.

When they came out

Mistral Large 4 High arrived on October 29, 2026, about 4 months after GPT-5.6 Luna Max (July 9, 2026).

Running costs

GPT-5.6 Luna Max lists $0.20 per million input tokens and $1.20 per million output tokens; Mistral Large 4 High lists $0.70 and $2.10. For output tokens, Mistral Large 4 High is 1.8 times the price of GPT-5.6 Luna Max.

For a sample workload of 1 million input tokens and 200,000 output tokens, GPT-5.6 Luna Max would cost $0.44 and Mistral Large 4 High $1.12, a difference of $0.68. For a month of coding-assistant traffic at 30 million input and 3 million output tokens, GPT-5.6 Luna Max would come in 65% lower: $9.60 for GPT-5.6 Luna Max against $27.30 for Mistral Large 4 High.

Cost per point of intelligence

Per point of intelligence index, GPT-5.6 Luna Max costs $0.019 of blended price per million tokens and Mistral Large 4 High $0.040. GPT-5.6 Luna Max delivers each index point for 53% less; whether Mistral Large 4 High's extra 2.3 points are worth the premium depends on how often your tasks need them.