The AI arena is free today

Open Superagent

DeepSeek-V3.2 (Thinking) vs Nemotron 3 Super (120B A12B)

DeepSeek-V3.2 (Thinking) leads the LLM Stats Score 32.6 to 26.8. Nemotron 3 Super (120B A12B) is 1.9x cheaper per token.

DeepSeek · NVIDIA · Updated for 2026

Which is better?

DeepSeek-V3.2 (Thinking) leads the overall LLM Stats Score 32.6 to 26.8, ranking #103 overall.

In the 10 individual benchmarks reported for both models, DeepSeek-V3.2 (Thinking) wins 8; this is a narrower head-to-head signal than the composite indexes.

On price, Nemotron 3 Super (120B A12B) is roughly 1.9x cheaper per token on a blended 3:1 input/output basis, which adds up quickly at production volume.

Nemotron 3 Super (120B A12B) also accepts a larger context window (262,144 input tokens), making it the stronger choice for long documents and large codebases.

Based on current LLM Stats indexes, shared benchmarks, pricing, and model metadata for 2026.

Choose DeepSeek-V3.2 (Thinking)

  • overall performance matters — it scores 32.6 and ranks #103 on LLM Stats
  • your work emphasizes reasoning and coding — it leads those capability indexes
  • you value its reported benchmark strengths — it wins 8 of 10 exact shared results

Choose Nemotron 3 Super (120B A12B)

  • cost matters — it's about 1.9x cheaper per token
  • you process long inputs — it offers a 262,144 token context window
  • you want the most recent training data — it shipped Mar 2026

At a glance

The differences that matter most.

DeepSeek-V3.2 (Thinking)
DeepSeek
Nemotron 3 Super (120B A12B)
NVIDIA
Core performance indexes
LLM Stats Score
32.6
#103
26.8
#146
Reasoning
32.6
#101
25.1
#152
Coding
22.9
#78
13.3
#136
Agents
11.2
#110
-0.4
#172
Cost, coverage & limits
Benchmark wins
8 of 10
2 of 10
Input price
$0.28 / M
$0.09 / M
Output price
$0.42 / M
$0.40 / M
Context window
131,072
262,144

Capability indexes

Additional strengths measured across groups of related public benchmarks

2 shared
Index
DeepSeek-V3.2 (Thinking)
Nemotron 3 Super (120B A12B)
Math
30.2#77
30.4#75
Tool use
10.0#129
2.9#165
Conservative TrueSkill rating · higher is betterHow scores work

Individual benchmarks

14 reported for DeepSeek-V3.2 (Thinking) · 23 for Nemotron 3 Super (120B A12B)

10 shared

DeepSeek-V3.2 (Thinking) outperforms in 8 benchmarks (AIME 2025, BrowseComp, Humanity's Last Exam, LiveCodeBench, MMLU-Pro, SWE-bench Multilingual, SWE-Bench Verified, Terminal-Bench 2.0), while Nemotron 3 Super (120B A12B) is better at 2 benchmarks (GPQA, HMMT 2025).

DeepSeek-V3.2 (Thinking) significantly outperforms across most benchmarks.

Fri Sep 11 2026 • llm-stats.com

Human preference

Blind head-to-head votes and playground preference scores

Pricing Analysis

Price comparison per million tokens

Nemotron 3 Super (120B A12B) costs less

For input processing, DeepSeek-V3.2 (Thinking) ($0.28/1M tokens) is 3.3x more expensive than Nemotron 3 Super (120B A12B) ($0.09/1M tokens).

For output processing, DeepSeek-V3.2 (Thinking) ($0.42/1M tokens) is 1.0x more expensive than Nemotron 3 Super (120B A12B) ($0.40/1M tokens).

In conclusion, DeepSeek-V3.2 (Thinking) is more expensive than Nemotron 3 Super (120B A12B).*

* Using a 3:1 ratio of input to output tokens

Lowest available price from all providers
Fri Sep 11 2026 • llm-stats.com
DeepSeek
DeepSeek-V3.2 (Thinking)
Input tokens$0.28
Output tokens$0.42
Best providerDeepSeek
NVIDIA
Nemotron 3 Super (120B A12B)
Input tokens$0.09
Output tokens$0.40
Best providerDeepinfra
Notice missing or incorrect data?Start an Issue→

Model Size

Parameter count comparison

565.0B diff

DeepSeek-V3.2 (Thinking) has 565.0B more parameters than Nemotron 3 Super (120B A12B), making it 470.8% larger.

DeepSeek
DeepSeek-V3.2 (Thinking)
685.0Bparameters
NVIDIA
Nemotron 3 Super (120B A12B)
120.0Bparameters
685.0B
DeepSeek-V3.2 (Thinking)
120.0B
Nemotron 3 Super (120B A12B)

Context Window

Maximum input and output token capacity

Nemotron 3 Super (120B A12B) accepts 262,144 input tokens compared to DeepSeek-V3.2 (Thinking)'s 131,072 tokens. Nemotron 3 Super (120B A12B) can generate longer responses up to 262,144 tokens, while DeepSeek-V3.2 (Thinking) is limited to 65,536 tokens.

DeepSeek
DeepSeek-V3.2 (Thinking)
Input131,072 tokens
Output65,536 tokens
NVIDIA
Nemotron 3 Super (120B A12B)
Input262,144 tokens
Output262,144 tokens
Fri Sep 11 2026 • llm-stats.com

License

Usage and distribution terms

DeepSeek-V3.2 (Thinking) is licensed under MIT, while Nemotron 3 Super (120B A12B) uses NVIDIA Open Model License Agreement .

License differences may affect how you can use these models in commercial or open-source projects.

DeepSeek-V3.2 (Thinking)

MIT

Open weights

Nemotron 3 Super (120B A12B)

NVIDIA Open Model License Agreement

Open weights

Release Timeline

When each model was launched

DeepSeek-V3.2 (Thinking) was released on 2025-12-01, while Nemotron 3 Super (120B A12B) was released on 2026-03-11.

Nemotron 3 Super (120B A12B) is 3 months newer than DeepSeek-V3.2 (Thinking).

DeepSeek-V3.2 (Thinking)

Dec 1, 2025

9 months ago

Nemotron 3 Super (120B A12B)

Mar 11, 2026

6 months ago

3mo newer

Knowledge Cutoff

When training data ends

Nemotron 3 Super (120B A12B) has a documented knowledge cutoff of 2025-06-01, while DeepSeek-V3.2 (Thinking)'s cutoff date is not specified.

We can confirm Nemotron 3 Super (120B A12B)'s training data extends to 2025-06-01, but cannot make a direct comparison without DeepSeek-V3.2 (Thinking)'s cutoff date.

DeepSeek-V3.2 (Thinking)

—

Nemotron 3 Super (120B A12B)

Jun 2025

Provider Availability

DeepSeek-V3.2 (Thinking) is available from DeepSeek. Nemotron 3 Super (120B A12B) is available from DeepInfra.

DeepSeek-V3.2 (Thinking)

deepseek logo
DeepSeek
Input Price:Input: $0.28/1MOutput Price:Output: $0.42/1M

Nemotron 3 Super (120B A12B)

deepinfra logo
Deepinfra
Input Price:Input: $0.09/1MOutput Price:Output: $0.40/1M
* Prices shown are per million tokens

Outputs Comparison

Notice missing or incorrect data?Start an Issue discussion→

Judge for yourself.

Run your own prompts against DeepSeek-V3.2 (Thinking) and Nemotron 3 Super (120B A12B) side-by-side, then vote on the output you prefer.

DeepSeek-V3.2 (Thinking)
✓ Preferred
Nemotron 3 Super (120B A12B)
Open in Playground

FAQ

Common questions about DeepSeek-V3.2 (Thinking) vs Nemotron 3 Super (120B A12B).

Which is better, DeepSeek-V3.2 (Thinking) or Nemotron 3 Super (120B A12B)?

DeepSeek-V3.2 (Thinking) leads the LLM Stats Score 32.6 to 26.8. DeepSeek-V3.2 (Thinking) is made by DeepSeek and Nemotron 3 Super (120B A12B) is made by NVIDIA. The best choice depends on your use case — compare their capability indexes, individual benchmarks, pricing, and limits above.

How does DeepSeek-V3.2 (Thinking) compare to Nemotron 3 Super (120B A12B) in benchmarks?

DeepSeek-V3.2 (Thinking) scores AIME 2025: 93.1%, HMMT 2025: 90.2%, MMLU-Pro: 85.0%, LiveCodeBench: 83.3%, GPQA: 82.4%. Nemotron 3 Super (120B A12B) scores HMMT 2025: 94.7%, RULER: 91.8%, AIME 2025: 90.2%, WMT24++: 86.7%, MMLU-Pro: 83.7%.

Is DeepSeek-V3.2 (Thinking) cheaper than Nemotron 3 Super (120B A12B)?

Nemotron 3 Super (120B A12B) is 3.3x cheaper for input tokens. DeepSeek-V3.2 (Thinking) costs $0.28/M input and $0.42/M output via deepseek. Nemotron 3 Super (120B A12B) costs $0.09/M input and $0.40/M output via deepinfra.

What are the context window sizes for DeepSeek-V3.2 (Thinking) and Nemotron 3 Super (120B A12B)?

DeepSeek-V3.2 (Thinking) supports 131K tokens and Nemotron 3 Super (120B A12B) supports 262K tokens. A larger context window lets you process longer documents, conversations, or codebases in a single request.

What are the main differences between DeepSeek-V3.2 (Thinking) and Nemotron 3 Super (120B A12B)?

Key differences include LLM Stats Score (32.6 vs 26.8), context window (131K vs 262K), input pricing ($0.28 vs $0.09/M), licensing (MIT vs NVIDIA Open Model License Agreement ). See the full comparison above for benchmark-by-benchmark results.

Who makes DeepSeek-V3.2 (Thinking) and Nemotron 3 Super (120B A12B)?

DeepSeek-V3.2 (Thinking) is developed by DeepSeek and Nemotron 3 Super (120B A12B) is developed by NVIDIA.

Related comparisons