The AI arena is free today

Open Superagent

DeepSeek-V3.2 (Thinking) vs Step-3.5-Flash

Step-3.5-Flash leads the LLM Stats Score 37.3 to 33.0. Step-3.5-Flash is 1.8x cheaper per token.

DeepSeek · StepFun · Updated for 2026

Which is better?

Step-3.5-Flash leads the overall LLM Stats Score 37.3 to 33.0, ranking #61 overall.

In the 4 individual benchmarks reported for both models, Step-3.5-Flash wins 4; this is a narrower head-to-head signal than the composite indexes.

On price, Step-3.5-Flash is roughly 1.8x cheaper per token on a blended 3:1 input/output basis, which adds up quickly at production volume.

DeepSeek-V3.2 (Thinking) also accepts a larger context window (131,072 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)

  • you process long inputs — it offers a 131,072 token context window

Choose Step-3.5-Flash

  • overall performance matters — it scores 37.3 and ranks #61 on LLM Stats
  • you value its reported benchmark strengths — it wins 4 of 4 exact shared results
  • cost matters — it's about 1.8x cheaper per token
  • you want the most recent training data — it shipped Feb 2026

At a glance

The differences that matter most.

DeepSeek-V3.2 (Thinking)
DeepSeek
Step-3.5-Flash
StepFun
Core performance indexes
LLM Stats Score
33.0
#90
37.3
#61
Reasoning
33.0
#90
37.3
#62
Coding
22.9
#69
20.4
#88
Agents
11.1
#97
15.3
#74
Cost, coverage & limits
Benchmark wins
0 of 4
4 of 4
Input price
$0.28 / M
$0.10 / M
Output price
$0.42 / M
$0.40 / M
Context window
131,072
65,536

Capability indexes

Additional strengths measured across groups of related public benchmarks

2 shared
Index
DeepSeek-V3.2 (Thinking)
Step-3.5-Flash
Math
30.6#70
33.8#50
Tool use
9.4#124
12.9#93
Conservative TrueSkill rating · higher is betterHow scores work

Individual benchmarks

14 reported for DeepSeek-V3.2 (Thinking) · 7 for Step-3.5-Flash

4 shared

DeepSeek-V3.2 (Thinking) outperforms in 0 benchmarks, while Step-3.5-Flash is better at 4 benchmarks (AIME 2025, BrowseComp, SWE-Bench Verified, Terminal-Bench 2.0).

Step-3.5-Flash significantly outperforms across most benchmarks.

Tue Sep 01 2026 • llm-stats.com

Human preference

Blind head-to-head votes and playground preference scores

Pricing Analysis

Price comparison per million tokens

Step-3.5-Flash costs less

For input processing, DeepSeek-V3.2 (Thinking) ($0.28/1M tokens) is 2.8x more expensive than Step-3.5-Flash ($0.10/1M tokens).

For output processing, DeepSeek-V3.2 (Thinking) ($0.42/1M tokens) is 1.0x more expensive than Step-3.5-Flash ($0.40/1M tokens).

In conclusion, DeepSeek-V3.2 (Thinking) is more expensive than Step-3.5-Flash.*

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

Lowest available price from all providers
Tue Sep 01 2026 • llm-stats.com
DeepSeek
DeepSeek-V3.2 (Thinking)
Input tokens$0.28
Output tokens$0.42
Best providerDeepSeek
StepFun
Step-3.5-Flash
Input tokens$0.10
Output tokens$0.40
Best providerStepFun
Notice missing or incorrect data?Start an Issue→

Model Size

Parameter count comparison

489.0B diff

DeepSeek-V3.2 (Thinking) has 489.0B more parameters than Step-3.5-Flash, making it 249.5% larger.

DeepSeek
DeepSeek-V3.2 (Thinking)
685.0Bparameters
StepFun
Step-3.5-Flash
196.0Bparameters
685.0B
DeepSeek-V3.2 (Thinking)
196.0B
Step-3.5-Flash

Context Window

Maximum input and output token capacity

DeepSeek-V3.2 (Thinking) accepts 131,072 input tokens compared to Step-3.5-Flash's 65,536 tokens. DeepSeek-V3.2 (Thinking) can generate longer responses up to 65,536 tokens, while Step-3.5-Flash is limited to 8,192 tokens.

DeepSeek
DeepSeek-V3.2 (Thinking)
Input131,072 tokens
Output65,536 tokens
StepFun
Step-3.5-Flash
Input65,536 tokens
Output8,192 tokens
Tue Sep 01 2026 • llm-stats.com

License

Usage and distribution terms

DeepSeek-V3.2 (Thinking) is licensed under MIT, while Step-3.5-Flash uses Apache 2.0.

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

DeepSeek-V3.2 (Thinking)

MIT

Open weights

Step-3.5-Flash

Apache 2.0

Open weights

Release Timeline

When each model was launched

DeepSeek-V3.2 (Thinking) was released on 2025-12-01, while Step-3.5-Flash was released on 2026-02-02.

Step-3.5-Flash is 2 months newer than DeepSeek-V3.2 (Thinking).

DeepSeek-V3.2 (Thinking)

Dec 1, 2025

9 months ago

Step-3.5-Flash

Feb 2, 2026

7 months ago

2mo newer

Knowledge Cutoff

When training data ends

Neither model specifies a knowledge cutoff date.

Unable to compare the recency of their training data.

No cutoff dates available

Provider Availability

DeepSeek-V3.2 (Thinking) is available from DeepSeek. Step-3.5-Flash is available from StepFun.

DeepSeek-V3.2 (Thinking)

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

Step-3.5-Flash

stepfun logo
StepFun
Input Price:Input: $0.10/1MOutput 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 Step-3.5-Flash side-by-side, then vote on the output you prefer.

DeepSeek-V3.2 (Thinking)
✓ Preferred
Step-3.5-Flash
Open in Playground

FAQ

Common questions about DeepSeek-V3.2 (Thinking) vs Step-3.5-Flash.

Which is better, DeepSeek-V3.2 (Thinking) or Step-3.5-Flash?

Step-3.5-Flash leads the LLM Stats Score 37.3 to 33.0. DeepSeek-V3.2 (Thinking) is made by DeepSeek and Step-3.5-Flash is made by StepFun. 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 Step-3.5-Flash 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%. Step-3.5-Flash scores AIME 2025: 97.3%, Tau-bench: 88.2%, LiveCodeBench v6: 86.4%, IMO-AnswerBench: 85.4%, SWE-Bench Verified: 74.4%.

Is DeepSeek-V3.2 (Thinking) cheaper than Step-3.5-Flash?

Step-3.5-Flash is 2.8x cheaper for input tokens. DeepSeek-V3.2 (Thinking) costs $0.28/M input and $0.42/M output via deepseek. Step-3.5-Flash costs $0.10/M input and $0.40/M output via stepfun.

What are the context window sizes for DeepSeek-V3.2 (Thinking) and Step-3.5-Flash?

DeepSeek-V3.2 (Thinking) supports 131K tokens and Step-3.5-Flash supports 66K 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 Step-3.5-Flash?

Key differences include LLM Stats Score (33.0 vs 37.3), context window (131K vs 66K), input pricing ($0.28 vs $0.10/M), licensing (MIT vs Apache 2.0). See the full comparison above for benchmark-by-benchmark results.

Who makes DeepSeek-V3.2 (Thinking) and Step-3.5-Flash?

DeepSeek-V3.2 (Thinking) is developed by DeepSeek and Step-3.5-Flash is developed by StepFun.

Related comparisons