Mistral logo

Pixtral Large

Retired May 2026

A 124B multimodal model from Mistral AI built on Mistral Large 2, with weights under the Mistral Research License, handling many high-resolution images per input.

Params
124B
Context
128K tokens
Max output
33K
License
Mistral Research License
Visit Mistral Compare vs other LLMs →

Estimated cost to self-host

Pixtral Large needs about 82 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× RTX 3060 at about $461 a month.

Single node ≤ 8
What it costs to self-host Pixtral Large at each precision: the memory it needs, the cheapest rental that holds it, that rental's month and the API volume it pays off at
Precision Memory Cheapest fit (32K context) Cost /mo
4-bit INT4 / FP4 82 GB Nvidia logo 8× RTX 3060 $461
8-bit FP8 / INT8 138 GB Nvidia logo 8× RTX 3090 $1,440
16-bit FP16 / BF16 262 GB Nvidia logo 8× RTX A6000 $3,283

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.

Capabilities

What Pixtral Large accepts and can do, as published by Mistral.

  • Text Accepts and generates natural-language text.
  • Image Accepts images as input alongside text.
  • Function calling Connect to external tools, APIs, and systems.
  • Structured output Return responses in structured formats like JSON.

Common questions

What is Pixtral Large good for?

Several high-resolution images in one input, built on Mistral Large 2. Open weights under the Mistral Research License, servable on a single high-memory GPU.

When is Pixtral Large not a good fit?

Any hosted use: Mistral retired it from its API in May 2026. The Mistral Research License is not a commercial licence, so check its terms before self-hosting for a business.

What is the cheapest way to run Pixtral Large?

No provider we track hosts it; self-hosting on the cheapest rental that fits (8x RTX 3060, $0.64 per hour) costs about $461 a month.

Can I self-host Pixtral Large?

Yes. The weights are released under the Mistral Research License. At 4-bit it needs about 82 GB of GPU memory, which starts at roughly $461 a month on the cheapest rental that fits.

More from Mistral

  • Ministral 3 14B Open weights 256K context
    From $0.20 / $0.20 /1M in / out at the cheapest provider
  • Ministral 3 3B Open weights 256K context
    From $0.10 / $0.10 /1M in / out at the cheapest provider
  • Ministral 3 8B Open weights 256K context
    From $0.15 / $0.15 /1M in / out at the cheapest provider
  • Mistral Large 3 Open weights 256K context
    From $0.50 / $1.50 /1M in / out at the cheapest provider
  • Mistral Medium 3.5 Open weights 256K context
    From $1.50 / $7.50 /1M in / out at the cheapest provider
  • Mistral Small 4 Open weights 256K context
    From $0.15 / $0.60 /1M in / out at the cheapest provider
  • Codestral 128K context
    From $0.30 / $0.90 /1M in / out at the cheapest provider
  • Devstral Medium 128K context
All Mistral models