DeepSeek

DeepSeek V3.2 Exp

Type
language model
Context
164K tokens
Max output
66K
Released
29 September 2025
Knowledge cutoff
January 2025
API string
deepseek-ai/deepseek-v3.2-exp
Weights
open

Prices by provider

Provider Input, $ per 1M Output, $ per 1M In our data since
OpenRouter openrouter/deepseek/deepseek-v3.2-exp $0.2 $0.4 31 Jul 2026
ZenMux $0.22 $0.33 31 Jul 2026
novita-ai $0.27 $0.41 31 Jul 2026
meganova $0.27 $0.4 31 Jul 2026
Kilo Code $0.27 $0.41 31 Jul 2026
Novita AI $0.27 $0.41 31 Jul 2026
OpenRouter deepseek/deepseek-v3.2-exp $0.27 $0.41 31 Jul 2026
SiliconFlow $0.27 $0.41 31 Jul 2026
NanoGPT $0.28 $0.42 31 Jul 2026
alibaba-cn $0.287 $0.431 31 Jul 2026

Only the base tier and only per-token prices are shown. Batch, discounted and cached rates, as well as prices per image or per second of video, do not go into this table: they cannot stand in the same column as a price per million tokens.

A single provider appears several times if it sells this model under different identifiers and at different prices — for example, at different weight precision. The identifier is shown next to the name. Identical offers that arrived from two sources under different spellings are merged into one row.

The date in the last column is the day this price first entered our collection. The price may well be older: before that day we simply were not recording it. It has not changed since — otherwise a new row with a new date would stand in its place.

Measurement results

Task set Result Run conditions Measured by
Arena Score in Spanish 1,441.73 1,408–1,475
Arena Score in English 1,439.48 1,431–1,448
Arena Score in French 1,436.53 1,394–1,479
Arena Score, programming 1,433.29 1,422–1,445
Arena Score, hard prompts 1,430.02 1,422–1,438
Arena Score, mathematics 1,424.1 1,404–1,445
Arena Score, multi-turn dialogue 1,423.37 1,410–1,436
Arena Score, overall 1,422.79 1,416–1,429
Arena Score, long queries 1,418.2 1,407–1,429
Arena Score, creative writing 1,403.82 1,389–1,418
Arena Score in Russian 1,403.49 1,384–1,423
Arena Score, instruction following 1,402.28 1,392–1,413
Arena Score, expert questions 1,400.8 1,377–1,424
Arena Score, web development 1,271.78 1,258–1,285
Fiction.LiveBench — holding a long context 83.3 % effort: high Fiction.live leaderboard
Aider Polyglot — code edits in six languages 74.2 % · with a tuned harness Aider LLM Leaderboards
The Agent Company — work tasks in an office environment 42.9 % TheAgentCompany leaderboard