Ad

SRExpert

Stop juggling 6 Kubernetes tools.

Monitoring, alerts, security, Helm deploys and AI ops in one platform.

Start a 30-day free trial

mistralai

Mistral raises prices on two models, cuts Small 3.2 24B

362 words

Mistral AI changed OpenRouter prices on three models on 2026-11-18: Mistral Medium 3 and Mistral Medium 3.1 rose 10.0 percent on input and output, while Mistral Small 3.2 24B Instruct fell 25.0 percent on input and 33.3 percent on output.

OpenRouter price, per 1M tokens

Measured on November 18, 2026

Mistral: Mistral Medium 3
Input /M tokens
$0.44
Output /M tokens
$2.20

Price changes in the last 90 days

  1. input $0.40 → $0.44, output $2.00 → $2.20

Full price history

Mistral: Mistral Medium 3.1
Input /M tokens
$0.44
Output /M tokens
$2.20

Price changes in the last 90 days

  1. input $0.40 → $0.44, output $2.00 → $2.20

Full price history

Mistral: Mistral Small 3.2 24B
Input /M tokens
$0.075
Output /M tokens
$0.20

Price changes in the last 90 days

  1. input $0.10 → $0.075, output $0.30 → $0.20
  2. input $0.075 → $0.10, output $0.20 → $0.30
  3. input $0.094 → $0.075, output $0.25 → $0.20
  4. input $0.075 → $0.094, output $0.20 → $0.25
  5. input $0.094 → $0.075, output $0.25 → $0.20
  6. input $0.075 → $0.094, output $0.20 → $0.25

Full price history

Mistral Medium 3 and Medium 3.1

Mistral Medium 3 now costs $0.44 per million input tokens and $2.2 per million output tokens, up from $0.4 and $2.0 on 2026-11-17. Mistral Medium 3.1 carries the same new prices and the same 10.0 percent increase. Both models have a 131072 token context window and are not open weights.

Mistral Small 3.2 24B Instruct

Mistral Small 3.2 24B Instruct now costs $0.075 per million input tokens and $0.2 per million output tokens, down from $0.1 and $0.3 on 2026-11-17. The model has 24.0 billion parameters, a 256000 token context window, and open weights. Its price history spans 89 days from 2026-08-22 with six changes; today's input price matches the record low first seen on 2026-08-22, after the previous change on 2026-11-10 set input at $0.1.

Official announcement

mistralai announced mistralai/Mistral-Small-3.2-24B-Instruct-2506. According to mistralai, Mistral-Small-3.2-24B-Instruct-2506 is a minor update of Mistral-Small-3.1-24B-Instruct-2503. According to mistralai, Small-3.2 is better at following precise instructions. According to mistralai, Small-3.2 produces less infinite generations or repetitive answers. According to mistralai, Small-3.2's function calling template is more robust. According to mistralai, Small 3.2 24B Instruct scores 65.33% on Wildbench v2. According to mistralai, Small 3.2 24B Instruct scores 43.1% on Arena Hard v2. According to mistralai, Small 3.2 24B Instruct scores 84.78% on IF (Internal; accuracy). According to mistralai, Small 3.2 reduces infinite generations by 2x on challenging, long and repetitive prompts. According to mistralai, Small 3.2 24B Instruct has an Infinite Generations (Internal) rate of 1.29%. According to mistralai, the model can be used with vllm and transformers. According to mistralai, running Mistral-Small-3.2-24B-Instruct-2506 on GPU requires ~55 GB of GPU RAM in bf16 or fp16. According to mistralai, the post recommends using a relatively low temperature, such as temperature=0.15.

  • Mistral Medium 3 and Mistral Medium 3.1 input and output prices each rose 10.0 percent to $0.44 and $2.2 per million tokens.
Mistral raises prices on two models, cuts Small 3.2 24B