token.app › Compare › Phi 4 vs Grok 4.6
Phi 4 vs Grok 4.6
Grok 4.6 leads on 2 of 2 shared benchmarks. Phi 4 is 34× cheaper per token on a blended 3:1 basis. Grok 4.6 accepts the larger context window (500K).
Head to head
Blended price uses a 3:1 input:output mix
| Phi 4 | Grok 4.6 | |
|---|---|---|
| Provider | Microsoft | SpaceXAI |
| Input $/1M | $0.070 | $2.00 |
| Output $/1M | $0.140 | $6.00 |
| Blended $/1M | $0.088 | $3.00 |
| Context window | 16K | 500K |
| Max output | 15K | 450K |
| Released | Jan 2025 | Aug 2026 |
| Licence | Open weights | Proprietary |
| GPQA Diamond | 56.1% | 94.0% |
| OTIS Mock AIME | 13.8% | 99.2% |
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 Phi 4 if…
- Costs less per token — $0.088 vs $3.00 blended
- Open weights — self-hostable
Choose Grok 4.6 if…
- Higher GPQA Diamond (94.0% vs 56.1%)
- Higher OTIS Mock AIME (99.2% vs 13.8%)
- Larger context window (500K)
- Newer model (released Aug 2026)
FAQ
Which is better, Phi 4 or Grok 4.6?
Across the 2 benchmarks both models have independently-run scores for, Phi 4 leads on 0 and Grok 4.6 leads on 2. Benchmarks are one input — price and context window are on this page too.
Is Phi 4 cheaper than Grok 4.6?
Phi 4 is cheaper: $0.088 versus $3.00 per million tokens blended at 3:1 input:output. Input alone: $0.070 vs $2.00. Output alone: $0.140 vs $6.00.
What context windows do Phi 4 and Grok 4.6 have?
Phi 4: 16,384 tokens. Grok 4.6: 500,000 tokens.
Who makes Phi 4 and Grok 4.6?
Phi 4 is made by Microsoft. Grok 4.6 is made by SpaceXAI.