token.app › Compare › Gemma 3 27B vs Phi 4

Gemma 3 27B vs Phi 4

Gemma 3 27B and Phi 4 split the 2 shared benchmarks evenly. Phi 4 is 2.0× cheaper per token on a blended 3:1 basis. Gemma 3 27B accepts the larger context window (262K).

Head to head
Blended price uses a 3:1 input:output mix
Gemma 3 27BPhi 4
ProviderGoogleMicrosoft
Input $/1M$0.080$0.070
Output $/1M$0.450$0.140
Blended $/1M$0.172$0.088
Context window262K16K
Max output131K16K
ReleasedMar 2025Jan 2025
LicenceOpen weightsOpen weights
GPQA Diamond 43.9% 56.1%
OTIS Mock AIME 22.2% 13.8%
Benchmark scores independently run and published by Epoch AI, 'AI Benchmarking Hub', used under CC BY 4.0. Only benchmarks with a score for BOTH models are compared.
Choose Gemma 3 27B if…
  • Higher OTIS Mock AIME (22.2% vs 13.8%)
  • Larger context window (262K)
  • Newer model (released Mar 2025)
Choose Phi 4 if…
  • Costs less per token — $0.088 vs $0.172 blended
  • Higher GPQA Diamond (56.1% vs 43.9%)
FAQ
Which is better, Gemma 3 27B or Phi 4?
Across the 2 benchmarks both models have independently-run scores for, Gemma 3 27B leads on 1 and Phi 4 leads on 1. Benchmarks are one input — price and context window are on this page too.
Is Gemma 3 27B cheaper than Phi 4?
Phi 4 is cheaper: $0.088 versus $0.172 per million tokens blended at 3:1 input:output. Input alone: $0.080 vs $0.070. Output alone: $0.450 vs $0.140.
What context windows do Gemma 3 27B and Phi 4 have?
Gemma 3 27B: 262,144 tokens. Phi 4: 16,384 tokens.
Who makes Gemma 3 27B and Phi 4?
Gemma 3 27B is made by Google. Phi 4 is made by Microsoft.
Related comparisons