Compare / DeepSeek-V3.2 vs DeepSeek-V3.2-Exp

DeepSeek-V3.2 vs DeepSeek-V3.2-Exp

Which is cheaper, which is more capable, and which does better at each kind of work, from 19 results on 4 sources that measured both models.

DeepSeekDeepSeek-V3.2 Released 1 Dec 2025 · $0.28 in · $0.42 out
DeepSeekDeepSeek-V3.2-Exp Released 29 Sep 2025 · $0.27 in · $0.41 out
DeepSeek-V3.2 wins 97 too close to callDeepSeek-V3.2-Exp wins 3

Of the 19 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).

DeepSeek-V3.2 or DeepSeek-V3.2-Exp? The short answer

DeepSeek-V3.2 is the newer version at about the same price, and it scores higher (Intelligence Index 21.5 against 16.6): upgrade.

  • They cost about the same: $0.32 and $0.30 per million tokens (three input to one output).
  • DeepSeek-V3.2 scores higher on the Artificial Analysis Intelligence Index: 21.5 against 16.6.
  • DeepSeek-V3.2 is newer: released 1 Dec 2025, against 29 Sep 2025 for DeepSeek-V3.2-Exp.
  • On third-party benchmarks, on the results both models have: DeepSeek-V3.2 does better for reasoning and knowledge and writing; DeepSeek-V3.2-Exp for sticking to the facts.

Head to head

21.5
Intelligence Index
16.6
$0.32
Price per million tokens, blendedlower is better
$0.30
163,840
Context window, tokens
163,840

Every fact side by side

DeepSeek-V3.2DeepSeek-V3.2-Exp
Price per million tokens$0.28 in · $0.42 out$0.27 in · $0.41 out
Blended (3 in : 1 out)$0.32$0.30
Cheapest host, blended$0.23 (GMICloud, FP8)–
Intelligence Index21.516.6
Spring Prompt overall––
Output speed, tokens a second––
Context window163,840 tokens163,840 tokens
Released1 Dec 2025 (newer)29 Sep 2025
WeightsOpenOpen
Developer based inChinaChina

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 93 of 159 76 of 159
Reasoning and knowledgeHard questions across science, maths and general knowledge 93 of 187 111 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 65 of 169 71 of 169
Sticking to the factsSummarising without inventing things, and factual answers 58 of 139 40 of 139

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureDeepSeek-V3.2DeepSeek-V3.2-Exp
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 106 · 105 of 387 measured 1,425 1,425thinking reasoning
Overall · Text factualityArena rating, higher is better 85 · 66 of 163 measured 1,437 1,446thinking reasoning
Business, management and finance · TextArena rating, higher is better 102 · 104 of 380 measured 1,423thinking reasoning 1,422
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 102 · 125 of 249 measured 21.5thinking reasoning 16.6thinking reasoning
GPQA Diamond% of questions, higher is better 90 · 111 of 232 measured 84.0%thinking reasoning 79.7%thinking reasoning
Humanity's Last Exam% of questions, higher is better 101 · 134 of 250 measured 24.6%thinking reasoning 14.9%thinking reasoning
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 54 · 64 of 230 measured 57.3thinking reasoning 54.0thinking reasoning
Requested-length error · UGI% off the requested word count, lower is better 151 · 147 of 230 measured 26.0%no reasoning 25.0%thinking reasoning
Style adherence · UGIscore from 0 to 1, higher is better 55 · 113 of 230 measured 0.37thinking reasoning 0.35no reasoning
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 27 · 15 of 104 measured 6.3% 5.3%
Answer rate · Vectara% of documents, higher is better 94 · 85 of 104 measured 92.6% 96.6%
Show all 19 results (8 more)
MeasureDeepSeek-V3.2DeepSeek-V3.2-Exp
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results
Creative writing · TextArena rating, higher is better 93 · 79 of 385 measured 1,401 1,410
Expert prompts · TextArena rating, higher is better 94 · 99 of 338 measured 1,452thinking reasoning 1,449thinking reasoning
Instruction following · TextArena rating, higher is better 94 · 98 of 387 measured 1,420 1,416thinking reasoning
Writing, literature and language · TextArena rating, higher is better 91 · 90 of 386 measured 1,410 1,411
Artificial Analysis reported by Artificial Analysis · 4 results
IFBench% of instructions, higher is better 79 · 93 of 195 measured 60.7%thinking reasoning 54.2%thinking reasoning
Long-context reasoning (AA-LCR)% of questions, higher is better 98 · 102 of 241 measured 73.3%thinking reasoning 72.3%thinking reasoning
Terminal-Bench Hard% of tasks, higher is better 47 · 65 of 190 measured 35.6%thinking reasoning 31.1%thinking reasoning
τ²-bench telecom% of tasks, higher is better 40 · 126 of 193 measured 90.6%thinking reasoning 33.9%no reasoning

Quick answers

Is DeepSeek-V3.2 better than DeepSeek-V3.2-Exp?

DeepSeek-V3.2 is the newer version at about the same price, and it scores higher (Intelligence Index 21.5 against 16.6): upgrade.

Which is cheaper, DeepSeek-V3.2 or DeepSeek-V3.2-Exp?

They cost about the same: $0.32 and $0.30 per million tokens (three input to one output).

Which is better for agents and tool use, DeepSeek-V3.2 or DeepSeek-V3.2-Exp?

DeepSeek-V3.2-Exp ranks 76 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 93 for DeepSeek-V3.2.

Which is better for reasoning and knowledge, DeepSeek-V3.2 or DeepSeek-V3.2-Exp?

DeepSeek-V3.2 ranks 93 of 187 models on sale for hard questions across science, maths and general knowledge, against 111 for DeepSeek-V3.2-Exp.

Which is better for writing, DeepSeek-V3.2 or DeepSeek-V3.2-Exp?

DeepSeek-V3.2 ranks 65 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 71 for DeepSeek-V3.2-Exp.

Which is better for sticking to the facts, DeepSeek-V3.2 or DeepSeek-V3.2-Exp?

DeepSeek-V3.2-Exp ranks 40 of 139 models on sale for summarising without inventing things, and factual answers, against 58 for DeepSeek-V3.2.

Run DeepSeek-V3.2 and DeepSeek-V3.2-Exp on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.