Compare / Gemma 3 27B vs Gemma 3 4B

Gemma 3 27B 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.

GoogleGemma 3 27B Released 11 Mar 2025 · $0.080 in · $0.35 out
GoogleGemma 3 4B Released 11 Mar 2025 · $0.040 in · $0.050 out
Gemma 3 27B wins 25Gemma 3 4B wins 4

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).

Gemma 3 27B or Gemma 3 4B? The short answer

They score about the same, so pick Gemma 3 4B: it costs 3.5 times less per token.

  • Gemma 3 4B is cheaper: $0.042 against $0.15 per million tokens, 3.5 times less.
  • They score about the same on the Artificial Analysis Intelligence Index (4.9 and 4.8).
  • On third-party benchmarks, on the results both models have: Gemma 3 27B does better for agents and tool use, reasoning and knowledge and writing; Gemma 3 4B for sticking to the facts.

Head to head

4.9
Intelligence Index
4.8
$0.15
Price per million tokens, blendedlower is better
$0.042
131,072
Context window, tokens
131,072

Every fact side by side

Gemma 3 27BGemma 3 4B
Price per million tokens$0.080 in · $0.35 out$0.040 in · $0.050 out
Blended (3 in : 1 out)$0.15$0.042
Cheapest host, blended$0.10 (DeepInfra, FP8)$0.042 (Parasail, BF16)
Intelligence Index4.94.8
Spring Prompt overall––
Output speed, tokens a second––
Context window131,072 tokens131,072 tokens
Released11 Mar 202511 Mar 2025
WeightsOpenOpen
Developer based inthe USthe 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 142 of 159 158 of 159
Reasoning and knowledgeHard questions across science, maths and general knowledge 180 of 187 181 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 106 of 169 157 of 169
Sticking to the factsSummarising without inventing things, and factual answers 92 of 139 32 of 139

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.

MeasureGemma 3 27BGemma 3 4B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 172 · 255 of 387 measured 1,365 1,303
Business, management and finance · TextArena rating, higher is better 162 · 208 of 380 measured 1,377 1,330
Creative writing · TextArena rating, higher is better 156 · 243 of 385 measured 1,349 1,275
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 245 · 246 of 249 measured 4.9 4.8
GPQA Diamond% of questions, higher is better 215 · 227 of 232 measured 42.8% 29.1%
Humanity's Last Exam% of questions, higher is better 215 · 194 of 250 measured 4.4% 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 50 · 74 of 81 measured 29.5%bfcl prompt 19.6%bfcl prompt
Irrelevance detection · BFCL% correct, higher is better 60 · 71 of 81 measured 73.7%bfcl prompt 53.9%bfcl prompt
Memory · BFCL% correct, higher is better 45 · 53 of 81 measured 13.6%bfcl prompt 8.6%bfcl prompt
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 88 · 185 of 230 measured 45.0 24.0
Requested-length error · UGI% off the requested word count, lower is better 25 · 33 of 230 measured 8.0% 9.0%
Style adherence · UGIscore from 0 to 1, higher is better 212 · 214 of 230 measured 0.29 0.29
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 35 · 28 of 104 measured 7.4% 6.4%
Answer rate · Vectara% of documents, higher is better 69 · 103 of 104 measured 98.8% 67.3%
Show all 29 results (15 more)
MeasureGemma 3 27BGemma 3 4B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Expert prompts · TextArena rating, higher is better 203 · 270 of 338 measured 1,337 1,260
Instruction following · TextArena rating, higher is better 185 · 270 of 387 measured 1,343 1,267
Writing, literature and language · TextArena rating, higher is better 163 · 237 of 386 measured 1,355 1,298
Artificial Analysis reported by Artificial Analysis · 7 results
Terminal-Bench 2.1% of tasks, higher is better 115 · 127 of 131 measured 4.5% 0.4%
τ-bench banking% of tasks, higher is better 119 · 121 of 121 measured 0.8% 0.4%
Artificial Analysis Coding Indexindex score, higher is better 120 · 129 of 130 measured 10.1 2.7
IFBench% of instructions, higher is better 173 · 187 of 195 measured 31.8% 28.3%
Long-context reasoning (AA-LCR)% of questions, higher is better 225 · 226 of 241 measured 7.3% 6.7%
Terminal-Bench Hard% of tasks, higher is better 162 · 180 of 190 measured 3.8% 0.8%
τ²-bench telecom% of tasks, higher is better 184 · 187 of 193 measured 10.5% 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 50 · 75 of 81 measured 10.8%bfcl prompt 0.4%bfcl prompt
Relevance detection · BFCL% correct, higher is better 36 · 1 of 81 measured 81.3%bfcl prompt 100.0%bfcl prompt
Single-turn calls (curated) · BFCL% correct, higher is better 25 · 74 of 81 measured 87.2%bfcl prompt 61.1%bfcl prompt
Single-turn calls (user-contributed) · BFCL% correct, higher is better 36 · 64 of 81 measured 74.5%bfcl prompt 60.8%bfcl prompt
Web search · BFCL% correct, higher is better 66 · 56 of 81 measured 0.0%bfcl prompt 1.0%bfcl prompt

Quick answers

Is Gemma 3 27B better than Gemma 3 4B?

They score about the same, so pick Gemma 3 4B: it costs 3.5 times less per token.

Which is cheaper, Gemma 3 27B or Gemma 3 4B?

Gemma 3 4B is cheaper: $0.042 against $0.15 per million tokens, 3.5 times less.

Which is better for agents and tool use, Gemma 3 27B or Gemma 3 4B?

Gemma 3 27B ranks 142 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, Gemma 3 27B or Gemma 3 4B?

Gemma 3 27B ranks 180 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, Gemma 3 27B or Gemma 3 4B?

Gemma 3 27B ranks 106 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, Gemma 3 27B or Gemma 3 4B?

Gemma 3 4B ranks 32 of 139 models on sale for summarising without inventing things, and factual answers, against 92 for Gemma 3 27B.

Run Gemma 3 27B 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.