Skip to content
Arena
Leaderboard

deepseek-v3.2

DeepSeekopen weightsinactiveThis model is no longer in battles and gets no new votes.

Licence
MIT
ID
deepseek-ai/deepseek-v3.2

deepseek-v3.2 no longer takes part in new comparisons. Its rating of 899 (±209) was set by the 9 votes cast while it took part. The rating's 95% interval is 690–1,108: it shows how precisely the votes have pinned the rating down so far. It won 14.3% of its decided comparisons (1 win, 6 losses, 2 ties). It is developed by DeepSeek. Its weights are open (licence: MIT).

At a glance

Rating
899±209
95% interval 690–1,108
Votes
9
14.3% of decided votes won

Model information

Note: The model is made available in the arena via the NVIDIA NIM API.

More about the model (opens in a new tab)

Statistics

Rating with 95% confidence interval

  1. EuroLLM-9B-Instructinactive
  2. llama-3.3-70b-instructinactive
  3. GaMS-9B-Instructinactive
  4. deepseek-v3.2inactive
  5. GaMS-27B-Instructinactive
  6. gpt-oss-20b
  7. Mistral-Small-24B-Instruct-2501inactive

The models just above and below it on the leaderboard, on one axis. The upright mark is the rating, the line its 95% interval.

Vote outcomes

  • Wins1
  • Ties2
  • Losses6
Total queries
27