Compare / Qwen3.5-397B-A17B vs DeepSeek-V4-Pro (0423)

Qwen3.5-397B-A17B vs DeepSeek-V4-Pro (0423)

Which is cheaper, which is more capable, and which does better at each kind of work, from 22 results on 3 sources that measured both models.

AlibabaQwen3.5-397B-A17B Released 13 Feb 2026 · $0.55 in · $3.50 out
DeepSeekDeepSeek-V4-Pro (0423) Released 24 Apr 2026 · $1.42 in · $2.83 out
Qwen3.5-397B-A17B wins 31 too close to callDeepSeek-V4-Pro (0423) wins 18

Of the 22 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 9 October 2026 (catalogue release 2026-10-09-dd620e778ac5). See also Alibaba vs DeepSeek, every model from both labs.

Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)? The short answer

Pick DeepSeek-V4-Pro (0423) if quality matters most: it scores higher (Intelligence Index 30.4 against 21.4). Pick Qwen3.5-397B-A17B if cost matters more: it costs 27% less per token.

  • Qwen3.5-397B-A17B is cheaper: $1.29 against $1.77 per million tokens, 27% less.
  • DeepSeek-V4-Pro (0423) scores higher on the Artificial Analysis Intelligence Index: 30.4 against 21.4.
  • DeepSeek-V4-Pro (0423) is newer: released 24 Apr 2026, against 13 Feb 2026 for Qwen3.5-397B-A17B.
  • On third-party benchmarks, on the results both models have: DeepSeek-V4-Pro (0423) does better for agents and tool use, reasoning and knowledge, writing and sticking to the facts.

Head to head

21.4
Intelligence Index
30.4
$1.29
Price per million tokens, blendedlower is better
$1.77
262,144
Context window, tokens
1,048,576

Every fact side by side

Qwen3.5-397B-A17BDeepSeek-V4-Pro (0423)
Price per million tokens$0.55 in · $3.50 out$1.42 in · $2.83 out
Blended (3 in : 1 out)$1.29$1.77
Cheapest host, blended$0.88 (Alibaba)$1.19 (StreamLake, FP8)
Intelligence Index21.430.4
Spring Prompt overall–33
Output speed, tokens a second90–
Context window262,144 tokens1,048,576 tokens
Released13 Feb 202624 Apr 2026 (newer)
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.

Decks from an analysisTurning a finished analysis into a deck you could present as it is not measured 18 of 19
Marketing planningPlanning a year of ad spend without overspending not measured 13 of 19
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 70 of 164 86 of 164
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 48 of 51 not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge 73 of 198 45 of 198
WritingWhat people prefer in blind comparisons, and judged writing quality 69 of 188 30 of 188
Sticking to the factsSummarising without inventing things, and factual answers 75 of 152 57 of 152
Speed under pressureGood decisions against a real clock (fast chess) not measured 24 of 24

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.

MeasureQwen3.5-397B-A17BDeepSeek-V4-Pro (0423)
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 80 · 53 of 387 measured 1,442 1,458
Overall · Text factualityArena rating, higher is better 70 · 49 of 163 measured 1,445 1,455
Business, management and finance · TextArena rating, higher is better 73 · 52 of 380 measured 1,446 1,460
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 97 · 49 of 243 measured 21.4no reasoning 30.4max reasoning
GPQA Diamond% of questions, higher is better 53 · 37 of 231 measured 89.3% 90.5%high reasoning
Humanity's Last Exam% of questions, higher is better 77 · 51 of 242 measured 29.0% 37.5%max reasoning
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 77 · 15 of 230 measured 50.1no reasoning 68.4thinking reasoning
Requested-length error · UGI% off the requested word count, lower is better 134 · 177 of 230 measured 23.0%no reasoning 34.0%thinking reasoning
Style adherence · UGIscore from 0 to 1, higher is better 164 · 76 of 230 measured 0.33no reasoning 0.36thinking reasoning
Show all 22 results (13 more)
MeasureQwen3.5-397B-A17BDeepSeek-V4-Pro (0423)
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results
Creative writing · TextArena rating, higher is better 86 · 39 of 385 measured 1,405 1,446
Expert prompts · TextArena rating, higher is better 67 · 64 of 338 measured 1,477 1,480
Instruction following · TextArena rating, higher is better 78 · 52 of 387 measured 1,434 1,452
Writing, literature and language · TextArena rating, higher is better 80 · 44 of 386 measured 1,421 1,451
Artificial Analysis reported by Artificial Analysis · 9 results
Terminal-Bench 2.1% of tasks, higher is better 61 · 41 of 130 measured 51.3% 64.8%high reasoning
Terminal-Bench 4.0% of tasks, higher is better 66 · 27 of 112 measured 0.0% 14.6%max reasoning
τ-bench banking% of tasks, higher is better 64 · 35 of 121 measured 13.4% 30.1%max reasoning
Artificial Analysis Coding Indexindex score, higher is better 61 · 40 of 131 measured 48.2 59.4max reasoning
IFBench% of instructions, higher is better 8 · 15 of 195 measured 78.8% 76.5%max reasoning
Long-context reasoning (AA-LCR)% of questions, higher is better 66 · 79 of 233 measured 77.3% 74.7%max reasoning
SciCode% of problems, higher is better 66 · 44 of 113 measured 44.8% 50.8%max reasoning
Terminal-Bench Hard% of tasks, higher is better 34 · 18 of 190 measured 40.9% 46.2%max reasoning
τ²-bench telecom% of tasks, higher is better 16 · 11 of 193 measured 95.6% 96.2%max reasoning

Quick answers

Is Qwen3.5-397B-A17B better than DeepSeek-V4-Pro (0423)?

Pick DeepSeek-V4-Pro (0423) if quality matters most: it scores higher (Intelligence Index 30.4 against 21.4). Pick Qwen3.5-397B-A17B if cost matters more: it costs 27% less per token.

Which is cheaper, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

Qwen3.5-397B-A17B is cheaper: $1.29 against $1.77 per million tokens, 27% less.

Which is better for agents and tool use, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

Qwen3.5-397B-A17B ranks 70 of 164 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 86 for DeepSeek-V4-Pro (0423).

Which is better for reasoning and knowledge, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

DeepSeek-V4-Pro (0423) ranks 45 of 198 models on sale for hard questions across science, maths and general knowledge, against 73 for Qwen3.5-397B-A17B.

Which is better for writing, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

DeepSeek-V4-Pro (0423) ranks 30 of 188 models on sale for what people prefer in blind comparisons, and judged writing quality, against 69 for Qwen3.5-397B-A17B.

Which is better for sticking to the facts, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

DeepSeek-V4-Pro (0423) ranks 57 of 152 models on sale for summarising without inventing things, and factual answers, against 75 for Qwen3.5-397B-A17B.

Which has the bigger context window, Qwen3.5-397B-A17B or DeepSeek-V4-Pro (0423)?

DeepSeek-V4-Pro (0423): 1,048,576 tokens, against 262,144 for Qwen3.5-397B-A17B.

Run Qwen3.5-397B-A17B and DeepSeek-V4-Pro (0423) on your own prompt

Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.

Try the playgroundTalk to us