Compare / Gemma 3 4B vs Phi-4

Gemma 3 4B vs Phi-4

Which is cheaper, which is more capable, and which does better at each kind of work, from 26 results on 5 sources that measured both models.

GoogleGemma 3 4B Released 11 Mar 2025 · $0.040 in · $0.050 out
MicrosoftPhi-4 Released 12 Dec 2024 · $0.070 in · $0.14 out
Gemma 3 4B wins 112 too close to callPhi-4 wins 13

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 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).

Gemma 3 4B or Phi-4? The short answer

They score about the same, so pick Gemma 3 4B: it costs 2.1 times less per token.

  • Gemma 3 4B is cheaper: $0.042 against $0.088 per million tokens, 2.1 times less.
  • Phi-4 scores higher on the Artificial Analysis Intelligence Index: 5.9 against 4.8.
  • Gemma 3 4B is newer: released 11 Mar 2025, against 12 Dec 2024 for Phi-4.
  • On third-party benchmarks, on the results both models have: Phi-4 does better for agents and tool use, reasoning and knowledge and sticking to the facts.

Head to head

4.8
Intelligence Index
5.9
$0.042
Price per million tokens, blendedlower is better
$0.088
131,072
Context window, tokens
16,384

Every fact side by side

Gemma 3 4BPhi-4
Price per million tokens$0.040 in · $0.050 out$0.070 in · $0.14 out
Blended (3 in : 1 out)$0.042$0.088
Cheapest host, blended$0.042 (Parasail, BF16)–
Intelligence Index4.85.9
Spring Prompt overall––
Output speed, tokens a second–45
Context window131,072 tokens16,384 tokens
Released11 Mar 2025 (newer)12 Dec 2024
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 158 of 159 115 of 159
Reasoning and knowledgeHard questions across science, maths and general knowledge 181 of 187 169 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 157 of 169 161 of 169
Sticking to the factsSummarising without inventing things, and factual answers 32 of 139 4 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 4BPhi-4
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 255 · 291 of 387 measured 1,303 1,256
Business, management and finance · TextArena rating, higher is better 208 · 282 of 380 measured 1,330 1,256
Creative writing · TextArena rating, higher is better 243 · 290 of 385 measured 1,275 1,210
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 246 · 237 of 249 measured 4.8 5.9
GPQA Diamond% of questions, higher is better 227 · 191 of 232 measured 29.1% 57.5%
Humanity's Last Exam% of questions, higher is better 194 · 229 of 250 measured 5.3% 3.8%
Berkeley Function Calling Leaderboard (BFCL) V4 reported by Berkeley Function Calling Leaderboard (BFCL) V4 · 3 results
Overall accuracy · BFCL% correct, higher is better 74 · 51 of 81 measured 19.6%bfcl prompt 28.8%bfcl prompt
Irrelevance detection · BFCL% correct, higher is better 71 · 16 of 81 measured 53.9%bfcl prompt 87.6%bfcl prompt
Memory · BFCL% correct, higher is better 53 · 24 of 81 measured 8.6%bfcl prompt 24.7%bfcl prompt
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 185 · 175 of 230 measured 24.0 25.7
Requested-length error · UGI% off the requested word count, lower is better 33 · 99 of 230 measured 9.0% 19.0%
Style adherence · UGIscore from 0 to 1, higher is better 214 · 122 of 230 measured 0.29 0.34
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 28 · 4 of 104 measured 6.4% 3.7%
Answer rate · Vectara% of documents, higher is better 103 · 100 of 104 measured 67.3% 80.7%
Show all 26 results (12 more)
MeasureGemma 3 4BPhi-4
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Expert prompts · TextArena rating, higher is better 270 · 261 of 338 measured 1,260 1,268
Instruction following · TextArena rating, higher is better 270 · 285 of 387 measured 1,267 1,245
Writing, literature and language · TextArena rating, higher is better 237 · 290 of 386 measured 1,298 1,229
Artificial Analysis reported by Artificial Analysis · 4 results
IFBench% of instructions, higher is better 187 · 192 of 195 measured 28.3% 23.5%
Long-context reasoning (AA-LCR)% of questions, higher is better 226 · 230 of 241 measured 6.7% 0.0%
Terminal-Bench Hard% of tasks, higher is better 180 · 162 of 190 measured 0.8% 3.8%
τ²-bench telecom% of tasks, higher is better 187 · 188 of 193 measured 5.0% 0.0%
Berkeley Function Calling Leaderboard (BFCL) V4 reported by Berkeley Function Calling Leaderboard (BFCL) V4 · 5 results
Multi-turn tasks · BFCL% correct, higher is better 75 · 64 of 81 measured 0.4%bfcl prompt 3.9%bfcl prompt
Relevance detection · BFCL% correct, higher is better 1 · 74 of 81 measured 100.0%bfcl prompt 50.0%bfcl prompt
Single-turn calls (curated) · BFCL% correct, higher is better 74 · 69 of 81 measured 61.1%bfcl prompt 69.6%bfcl prompt
Single-turn calls (user-contributed) · BFCL% correct, higher is better 64 · 66 of 81 measured 60.8%bfcl prompt 60.7%bfcl prompt
Web search · BFCL% correct, higher is better 56 · 44 of 81 measured 1.0%bfcl prompt 4.5%bfcl prompt

Quick answers

Is Gemma 3 4B better than Phi-4?

They score about the same, so pick Gemma 3 4B: it costs 2.1 times less per token.

Which is cheaper, Gemma 3 4B or Phi-4?

Gemma 3 4B is cheaper: $0.042 against $0.088 per million tokens, 2.1 times less.

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

Phi-4 ranks 115 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 4B or Phi-4?

Phi-4 ranks 169 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 4B or Phi-4?

Gemma 3 4B ranks 157 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 161 for Phi-4.

Which is better for sticking to the facts, Gemma 3 4B or Phi-4?

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

Which has the bigger context window, Gemma 3 4B or Phi-4?

Gemma 3 4B: 131,072 tokens, against 16,384 for Phi-4.

Run Gemma 3 4B and Phi-4 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.