Compare / Qwen3-32B vs Gemma 3 4B
Qwen3-32B vs Gemma 3 4B
Which is cheaper, which is more capable, and which does better at each kind of work, from 29 results on 5 sources that measured both models.
Of the 29 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). See also Alibaba vs Google, every model from both labs.
Qwen3-32B or Gemma 3 4B? The short answer
Pick Qwen3-32B if quality matters most: it scores higher (Intelligence Index 7.3 against 4.8). Pick Gemma 3 4B if cost matters more: it costs 4.8 times less per token.
- Gemma 3 4B is cheaper: $0.042 against $0.21 per million tokens, 4.8 times less.
- Qwen3-32B scores higher on the Artificial Analysis Intelligence Index: 7.3 against 4.8.
- Qwen3-32B is newer: released 28 Apr 2025, against 11 Mar 2025 for Gemma 3 4B.
- On third-party benchmarks, on the results both models have: Qwen3-32B does better for agents and tool use, reasoning and knowledge, writing and sticking to the facts.
Head to head
Every fact side by side
| Qwen3-32B | Gemma 3 4B | |
|---|---|---|
| Price per million tokens | $0.14 in · $0.40 out | $0.040 in · $0.050 out |
| Blended (3 in : 1 out) | $0.21 | $0.042 |
| Cheapest host, blended | $0.13 (DeepInfra, FP8) | $0.042 (Parasail, BF16) |
| Intelligence Index | 7.3 | 4.8 |
| Spring Prompt overall | – | – |
| Output speed, tokens a second | – | – |
| Context window | 131,072 tokens | 131,072 tokens |
| Released | 28 Apr 2025 (newer) | 11 Mar 2025 |
| Weights | Open | Open |
| Developer based in | China | the 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.
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.
| Measure | Qwen3-32B | Gemma 3 4B |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Overall · TextArena rating, higher is better 195 · 255 of 387 measured | 1,347 | 1,303 |
| Business, management and finance · TextArena rating, higher is better 203 · 208 of 380 measured | 1,340 | 1,330 |
| Creative writing · TextArena rating, higher is better 205 · 243 of 385 measured | 1,304 | 1,275 |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 213 · 246 of 249 measured | 7.3no reasoning | 4.8 |
| GPQA Diamond% of questions, higher is better 168 · 227 of 232 measured | 66.8%thinking reasoning | 29.1% |
| Humanity's Last Exam% of questions, higher is better 176 · 194 of 250 measured | 7.4%thinking reasoning | 5.3% |
| Berkeley Function Calling Leaderboard (BFCL) V4 reported by Berkeley Function Calling Leaderboard (BFCL) V4 · 3 results | ||
| Overall accuracy · BFCL% correct, higher is better 24 · 74 of 81 measured | 48.7%bfcl fc | 19.6%bfcl prompt |
| Irrelevance detection · BFCL% correct, higher is better 37 · 71 of 81 measured | 82.4%bfcl prompt | 53.9%bfcl prompt |
| Memory · BFCL% correct, higher is better 21 · 53 of 81 measured | 26.7%bfcl fc | 8.6%bfcl prompt |
| UGI Leaderboard reported by UGI Leaderboard · 3 results | ||
| Writing score · UGIscore out of 100, higher is better 134 · 185 of 230 measured | 33.0no reasoning | 24.0 |
| Requested-length error · UGI% off the requested word count, lower is better 99 · 33 of 230 measured | 19.0%no reasoning | 9.0% |
| Style adherence · UGIscore from 0 to 1, higher is better 175 · 214 of 230 measured | 0.32no reasoning | 0.29 |
| Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results | ||
| Hallucination rate · Vectara% of summaries, lower is better 24 · 28 of 104 measured | 5.9% | 6.4% |
| Answer rate · Vectara% of documents, higher is better 17 · 103 of 104 measured | 99.9% | 67.3% |
Show all 29 results (15 more)
| Measure | Qwen3-32B | Gemma 3 4B |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Expert prompts · TextArena rating, higher is better 159 · 270 of 338 measured | 1,398 | 1,260 |
| Instruction following · TextArena rating, higher is better 198 · 270 of 387 measured | 1,331 | 1,267 |
| Writing, literature and language · TextArena rating, higher is better 217 · 237 of 386 measured | 1,308 | 1,298 |
| Artificial Analysis reported by Artificial Analysis · 7 results | ||
| Terminal-Bench 2.1% of tasks, higher is better 112 · 127 of 131 measured | 5.2%thinking reasoning | 0.4% |
| τ-bench banking% of tasks, higher is better 102 · 121 of 121 measured | 5.4%thinking reasoning | 0.4% |
| Artificial Analysis Coding Indexindex score, higher is better 110 · 129 of 130 measured | 15.3thinking reasoning | 2.7 |
| IFBench% of instructions, higher is better 155 · 187 of 195 measured | 36.3%thinking reasoning | 28.3% |
| Long-context reasoning (AA-LCR)% of questions, higher is better 230 · 226 of 241 measured | 0.0%no reasoning | 6.7% |
| Terminal-Bench Hard% of tasks, higher is better 169 · 180 of 190 measured | 3.0%thinking reasoning | 0.8% |
| τ²-bench telecom% of tasks, higher is better 135 · 187 of 193 measured | 29.8%thinking reasoning | 5.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4 reported by Berkeley Function Calling Leaderboard (BFCL) V4 · 5 results | ||
| Multi-turn tasks · BFCL% correct, higher is better 17 · 75 of 81 measured | 47.9%bfcl fc | 0.4%bfcl prompt |
| Relevance detection · BFCL% correct, higher is better 8 · 1 of 81 measured | 93.8%bfcl fc | 100.0%bfcl prompt |
| Single-turn calls (curated) · BFCL% correct, higher is better 3 · 74 of 81 measured | 90.3%bfcl prompt | 61.1%bfcl prompt |
| Single-turn calls (user-contributed) · BFCL% correct, higher is better 3 · 64 of 81 measured | 82.0%bfcl fc | 60.8%bfcl prompt |
| Web search · BFCL% correct, higher is better 28 · 56 of 81 measured | 26.0%bfcl prompt | 1.0%bfcl prompt |
Compare them with other models
Quick answers
Is Qwen3-32B better than Gemma 3 4B?
Pick Qwen3-32B if quality matters most: it scores higher (Intelligence Index 7.3 against 4.8). Pick Gemma 3 4B if cost matters more: it costs 4.8 times less per token.
Which is cheaper, Qwen3-32B or Gemma 3 4B?
Gemma 3 4B is cheaper: $0.042 against $0.21 per million tokens, 4.8 times less.
Which is better for agents and tool use, Qwen3-32B or Gemma 3 4B?
Qwen3-32B ranks 116 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 158 for Gemma 3 4B.
Which is better for reasoning and knowledge, Qwen3-32B or Gemma 3 4B?
Qwen3-32B ranks 148 of 187 models on sale for hard questions across science, maths and general knowledge, against 181 for Gemma 3 4B.
Which is better for writing, Qwen3-32B or Gemma 3 4B?
Qwen3-32B ranks 128 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 157 for Gemma 3 4B.
Which is better for sticking to the facts, Qwen3-32B or Gemma 3 4B?
Qwen3-32B ranks 22 of 139 models on sale for summarising without inventing things, and factual answers, against 32 for Gemma 3 4B.
Run Qwen3-32B and Gemma 3 4B 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.