Model comparison
DeepSeek-R1-Distill-Llama-70B vs Phi-4
DeepSeek-R1-Distill-Llama-70B is the stronger model overall, scoring 37.8 to 31.2 on the Noometry Index.
Last verified . 12 shared benchmarks.
DeepSeek
MicrosoftSummary
- They share 12 benchmarks with published results for both. DeepSeek-R1-Distill-Llama-70B scores higher in 5 categories and Phi-4 in 1 category; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-R1-Distill-Llama-70B leads 36.0 to 20.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 51.4% for DeepSeek-R1-Distill-Llama-70B and 13.8% for Phi-4.
Side by side
| DeepSeek-R1-Distill-Llama-70B | Phi-4 | |
|---|---|---|
| Provider | DeepSeek | Microsoft |
| Noometry Index | 37.8 | 31.2 |
| Released | 2025-01-20 | 2024-12-11 |
| Weights | Open | Open |
| Context window | — | 128K |
| Max output | — | 4K |
| Input $ / M tokens | — | $0.07 |
| Output $ / M tokens | — | $0.14 |
| Results tracked | 13 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1-Distill-Llama-70B leads
DeepSeek-R1-Distill-Llama-70B: 36.8 (#202), Phi-4: 34.4 (#239)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| BigCodeBench Instruct | 35.3% | 45.5% |
| LiveBench Coding | 51.6% | 30.7% |
| BigCodeBench Complete | 49.9% | 55.4% |
| LMArena Coding | — | 1231 |
Agentic & Tool Use Not comparable
DeepSeek-R1-Distill-Llama-70B: —, Phi-4: 22.8 (#128)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
Reasoning DeepSeek-R1-Distill-Llama-70B leads
DeepSeek-R1-Distill-Llama-70B: 24.9 (#156), Phi-4: 17.7 (#291)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| LiveBench Reasoning | 67.6% | 47.8% |
| LiveBench Data Analysis | 55.9% | 45.2% |
| LiveBench | 54.5% | 41.6% |
| Kagi LLM Benchmark | 52.3% | — |
| Chess Puzzles | — | 1% |
| LMArena Hard Prompts | — | 1220 |
| Epoch Capabilities Index | — | 130.42 |
Math DeepSeek-R1-Distill-Llama-70B leads
DeepSeek-R1-Distill-Llama-70B: 36.0 (#176), Phi-4: 20.8 (#285)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 51.4% | 13.8% |
| LiveBench Math | 58.1% | 42% |
| MATH Level 5 | 89.9% | 64.9% |
| LMArena Math | — | 1246 |
Knowledge Phi-4 leads
DeepSeek-R1-Distill-Llama-70B: 30.7 (#225), Phi-4: 32.6 (#209)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| GPQA Diamond | 55.7% | 56.1% |
| Confabulations | — | 29.4% |
| Vectara Hallucination Rate | — | 3.7% |
| LMArena Expert | — | 1203 |
| MMLU | — | 84.8% |
Multilingual Not comparable
DeepSeek-R1-Distill-Llama-70B: —, Phi-4: 37.2 (#237)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| LMArena Non-English | — | 1197 |
| LMArena Chinese | — | 1212 |
| LMArena French | — | 1224 |
| LMArena German | — | 1222 |
| LMArena Japanese | — | 1158 |
| LMArena Korean | — | 1151 |
| LMArena Russian | — | 1209 |
| LMArena Spanish | — | 1234 |
Instruction Following DeepSeek-R1-Distill-Llama-70B leads
DeepSeek-R1-Distill-Llama-70B: 68.2 (#190), Phi-4: 60.4 (#251)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| LiveBench Instruction Following | 69.9% | 58.4% |
| LMArena Instruction Following | — | 1201 |
Long Context Not comparable
DeepSeek-R1-Distill-Llama-70B: —, Phi-4: 36.9 (#226)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| LMArena Longer Query | — | 1217 |
Writing & Preference DeepSeek-R1-Distill-Llama-70B leads
DeepSeek-R1-Distill-Llama-70B: 49.0 (#194), Phi-4: 40.5 (#244)
| Benchmark | DeepSeek-R1-Distill-Llama-70B | Phi-4 |
|---|---|---|
| LiveBench Language | 23.8% | 25.6% |
| LMArena Text | — | 1217 |
| LMArena Creative Writing | — | 1182 |
| Short-Story Creative Writing | — | 62.6% |
| LMArena Multi-Turn | — | 1206 |
Frequently asked questions
Is DeepSeek-R1-Distill-Llama-70B better than Phi-4?
DeepSeek-R1-Distill-Llama-70B is the stronger model overall, scoring 37.8 to 31.2 on the Noometry Index.
Is DeepSeek-R1-Distill-Llama-70B or Phi-4 better for coding?
DeepSeek-R1-Distill-Llama-70B scores higher on coding benchmarks: 36.8 versus 34.4 in the Noometry coding category.
How many benchmarks do DeepSeek-R1-Distill-Llama-70B and Phi-4 share?
12 benchmarks have published results for both models. DeepSeek-R1-Distill-Llama-70B has 13 scored results on Noometry and Phi-4 has 37.