Model comparison
Inkling vs Phi-4
Inkling is the stronger model overall, scoring 44.1 to 31.2 on the Noometry Index. Phi-4 costs 29× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Last verified . 21 shared benchmarks.
MicrosoftSummary
- They share 21 benchmarks with published results for both. Inkling scores higher in 9 categories and Phi-4 in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Inkling leads 65.2 to 40.5.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 88.9% for Inkling and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $1.87 / $4.68 for Inkling.
- Phi-4 accepts more context: 128K tokens versus 66K.
Side by side
| Inkling | Phi-4 | |
|---|---|---|
| Provider | Thinking Machines Lab | Microsoft |
| Noometry Index | 44.1 | 31.2 |
| Released | 2026-07-15 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 66K | 128K |
| Max output | 66K | 4K |
| Input $ / M tokens | $1.87 | $0.07 |
| Output $ / M tokens | $4.68 | $0.14 |
| Results tracked | 41 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Inkling: 34.5 (#234), Phi-4: 34.4 (#239)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| LMArena Coding | 1464 | 1231 |
| FrontierCode | 14% | — |
| LMArena WebDev | 1413 | — |
| FrontierSWE | 4.1% | — |
| SciCode | 47% | — |
| WeirdML | 32.3% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 946 | — |
Agentic & Tool Use Inkling leads
Inkling: 29.6 (#85), Phi-4: 22.8 (#128)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| APEX-Agents | 33.8% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| τ²-bench Banking | 25% | — |
| BALROG | — | 11.6% |
Reasoning Inkling leads
Inkling: 40.4 (#56), Phi-4: 17.7 (#291)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| Chess Puzzles | 21% | 1% |
| LMArena Hard Prompts | 1451 | 1220 |
| Epoch Capabilities Index | 148.54 | 130.42 |
| ARC-AGI-2 | 36.5% | — |
| SimpleBench | 50% | — |
| ARC-AGI-1 | 79.5% | — |
| CritPt | 5.4% | — |
| LiveBench Reasoning | — | 47.8% |
| DTBench | 87.5% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 37.6% | — |
| LiveBench | — | 41.6% |
Math Inkling leads
Inkling: 31.3 (#225), Phi-4: 20.8 (#285)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 13.8% |
| LMArena Math | 1479 | 1246 |
| FrontierMath (Tiers 1-3) | 33.3% | — |
| FrontierMath Tier 4 | 4.9% | — |
| ProofBench | 0% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
Knowledge Inkling leads
Inkling: 55.1 (#49), Phi-4: 32.6 (#209)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| GPQA Diamond | 88.3% | 56.1% |
| LMArena Expert | 1465 | 1203 |
| SimpleQA Verified | 40.3% | — |
| Confabulations | — | 29.4% |
| Vectara Hallucination Rate | — | 3.7% |
| MMLU | — | 84.8% |
Multilingual Inkling leads
Inkling: 54.0 (#52), Phi-4: 37.2 (#237)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| LMArena Non-English | 1434 | 1197 |
| LMArena Chinese | 1490 | 1212 |
| LMArena French | 1458 | 1224 |
| LMArena German | 1446 | 1222 |
| LMArena Japanese | 1429 | 1158 |
| LMArena Korean | 1404 | 1151 |
| LMArena Russian | 1429 | 1209 |
| LMArena Spanish | 1448 | 1234 |
Instruction Following Inkling leads
Inkling: 75.1 (#71), Phi-4: 60.4 (#251)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1426 | 1201 |
| LiveBench Instruction Following | — | 58.4% |
Long Context Inkling leads
Inkling: 43.8 (#86), Phi-4: 36.9 (#226)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1434 | 1217 |
Writing & Preference Inkling leads
Inkling: 65.2 (#51), Phi-4: 40.5 (#244)
| Benchmark | Inkling | Phi-4 |
|---|---|---|
| LMArena Text | 1441 | 1217 |
| LMArena Creative Writing | 1387 | 1182 |
| LMArena Multi-Turn | 1436 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench Creative Writing | 1611 | — |
| EQ-Bench 4 | 1226 | — |
| LiveBench Language | — | 25.6% |
Frequently asked questions
Is Inkling better than Phi-4?
Inkling is the stronger model overall, scoring 44.1 to 31.2 on the Noometry Index. Phi-4 costs 29× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Which is cheaper, Inkling or Phi-4?
Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Inkling lists at $1.87 and $4.68.
Is Inkling or Phi-4 better for coding?
They score almost the same on coding (34.5 vs 34.4); test both on your own repository before choosing.
Which has the bigger context window?
Phi-4 does, with 128K tokens against 66K.
How many benchmarks do Inkling and Phi-4 share?
21 benchmarks have published results for both models. Inkling has 41 scored results on Noometry and Phi-4 has 37.