Compare / Qwen3.5-122B-A10B vs Gemma 4 31B

Qwen3.5-122B-A10B vs Gemma 4 31B

Which is cheaper, which is more capable, and which does better at each kind of work, from 26 results on 4 sources that measured both models.

AlibabaQwen3.5-122B-A10B Released 24 Feb 2026 · $0.26 in · $2.08 out
GoogleGemma 4 31B Released 2 Apr 2026 · $0.14 in · $0.40 out
Qwen3.5-122B-A10B wins 92 too close to callGemma 4 31B wins 15

Of the 26 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 Google, every model from both labs.

Qwen3.5-122B-A10B or Gemma 4 31B? The short answer

Pick Qwen3.5-122B-A10B if quality matters most: it scores higher (Intelligence Index 17.7 against 14.7). Pick Gemma 4 31B if cost matters more: it costs 3.5 times less per token. Qwen3.5-122B-A10B is also faster.

  • Gemma 4 31B is cheaper: $0.21 against $0.72 per million tokens, 3.5 times less.
  • Qwen3.5-122B-A10B scores higher on the Artificial Analysis Intelligence Index: 17.7 against 14.7.
  • Qwen3.5-122B-A10B writes faster: 147 tokens a second against 43.
  • Gemma 4 31B is newer: released 2 Apr 2026, against 24 Feb 2026 for Qwen3.5-122B-A10B.
  • On third-party benchmarks, on the results both models have: Qwen3.5-122B-A10B does better for agents and tool use and reasoning and knowledge; Gemma 4 31B for writing and sticking to the facts.

Head to head

17.7
Intelligence Index
14.7
$0.72
Price per million tokens, blendedlower is better
$0.21
147
Output speed, tokens a second
43
262,144
Context window, tokens
262,144

Every fact side by side

Qwen3.5-122B-A10BGemma 4 31B
Price per million tokens$0.26 in · $2.08 out$0.14 in · $0.40 out
Blended (3 in : 1 out)$0.72$0.21
Cheapest host, blended$0.72 (Alibaba)$0.15 (DeepInfra, FP4)
Intelligence Index17.714.7
Spring Prompt overall––
Output speed, tokens a second14743
Context window262,144 tokens262,144 tokens
Released24 Feb 20262 Apr 2026 (newer)
WeightsOpenOpen
Developer based inChinathe US

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 95 of 164 101 of 164
Reasoning and knowledgeHard questions across science, maths and general knowledge 87 of 198 93 of 198
WritingWhat people prefer in blind comparisons, and judged writing quality 102 of 188 74 of 188
Sticking to the factsSummarising without inventing things, and factual answers 108 of 152 90 of 152

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.

MeasureQwen3.5-122B-A10BGemma 4 31B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 117 · 65 of 387 measured 1,416 1,453
Overall · Text factualityArena rating, higher is better 103 · 60 of 163 measured 1,428 1,450
Business, management and finance · TextArena rating, higher is better 105 · 66 of 380 measured 1,422 1,449
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 114 · 130 of 243 measured 17.7no reasoning 14.7
GPQA Diamond% of questions, higher is better 75 · 75 of 231 measured 85.7% 85.7%
Humanity's Last Exam% of questions, higher is better 92 · 97 of 242 measured 25.2% 23.6%
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 112 · 94 of 230 measured 39.5no reasoning 43.1channel-thought-prefill reasoning
Requested-length error · UGI% off the requested word count, lower is better 33 · 15 of 230 measured 9.0%thinking reasoning 5.0%
Style adherence · UGIscore from 0 to 1, higher is better 83 · 73 of 230 measured 0.36no reasoning 0.36channel-thought-prefill reasoning
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 72 · 35 of 104 measured 11.2% 7.4%
Answer rate · Vectara% of documents, higher is better 31 · 1 of 104 measured 99.8% 100.0%
Show all 26 results (15 more)
MeasureQwen3.5-122B-A10BGemma 4 31B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results
Creative writing · TextArena rating, higher is better 136 · 74 of 385 measured 1,366 1,420
Expert prompts · TextArena rating, higher is better 111 · 42 of 338 measured 1,442 1,498
Instruction following · TextArena rating, higher is better 115 · 49 of 387 measured 1,406 1,453
Writing, literature and language · TextArena rating, higher is better 125 · 63 of 386 measured 1,387 1,435
Artificial Analysis reported by Artificial Analysis · 11 results
Terminal-Bench 2.1% of tasks, higher is better 64 · 71 of 130 measured 47.6% 43.4%
Terminal-Bench 4.0% of tasks, higher is better 66 · 66 of 112 measured 0.0% 0.0%
τ-bench banking% of tasks, higher is better 58 · 61 of 121 measured 15.3% 14.8%
Artificial Analysis Coding Indexindex score, higher is better 65 · 70 of 131 measured 45.7 43.4
IFBench% of instructions, higher is better 22 · 24 of 195 measured 75.7% 75.6%
Long-context reasoning (AA-LCR)% of questions, higher is better 72 · 112 of 233 measured 76.3% 69.7%
Output speedtokens per second, higher is better 21 · 68 of 74 measured 147no reasoning 42.9no reasoning
SciCode% of problems, higher is better 79 · 63 of 113 measured 39.7% 45.5%
Terminal-Bench Hard% of tasks, higher is better 65 · 46 of 190 measured 31.1% 36.4%
Time to first answer tokenseconds, lower is better 29 · 20 of 74 measured 1.05no reasoning 0.83no reasoning
τ²-bench telecom% of tasks, higher is better 30 · 88 of 193 measured 93.6% 65.5%no reasoning

Quick answers

Is Qwen3.5-122B-A10B better than Gemma 4 31B?

Pick Qwen3.5-122B-A10B if quality matters most: it scores higher (Intelligence Index 17.7 against 14.7). Pick Gemma 4 31B if cost matters more: it costs 3.5 times less per token. Qwen3.5-122B-A10B is also faster.

Which is cheaper, Qwen3.5-122B-A10B or Gemma 4 31B?

Gemma 4 31B is cheaper: $0.21 against $0.72 per million tokens, 3.5 times less.

Which is faster, Qwen3.5-122B-A10B or Gemma 4 31B?

Qwen3.5-122B-A10B writes about 147 tokens a second on its usual API, against 43 for Gemma 4 31B (Artificial Analysis).

Which is better for agents and tool use, Qwen3.5-122B-A10B or Gemma 4 31B?

Qwen3.5-122B-A10B ranks 95 of 164 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 101 for Gemma 4 31B.

Which is better for reasoning and knowledge, Qwen3.5-122B-A10B or Gemma 4 31B?

Qwen3.5-122B-A10B ranks 87 of 198 models on sale for hard questions across science, maths and general knowledge, against 93 for Gemma 4 31B.

Which is better for writing, Qwen3.5-122B-A10B or Gemma 4 31B?

Gemma 4 31B ranks 74 of 188 models on sale for what people prefer in blind comparisons, and judged writing quality, against 102 for Qwen3.5-122B-A10B.

Which is better for sticking to the facts, Qwen3.5-122B-A10B or Gemma 4 31B?

Gemma 4 31B ranks 90 of 152 models on sale for summarising without inventing things, and factual answers, against 108 for Qwen3.5-122B-A10B.

Run Qwen3.5-122B-A10B and Gemma 4 31B 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